Claudexor

NPM · CLAUDEXOR

Route coding work across Claude Code, Codex, Cursor, and OpenCode with shared context.

Available components

npm · claudexor 89

0 this week 89 89 Trust /100

Last scanned 4 Nov 2026 (18 hours ago)

Scores higher than 93% of npm servers

Category Developer Tools

Trust breakdown (7 categories)

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. How we score → Why this is hard to score →

Supply Chain Security91

No malware found by supply-chain analysis.Pass
CVE check failed: a known high-severity CVE affects @modelcontextprotocol/server 2.2.0, reached via @claudexor/cli > @claudexor/mcp-server > @modelcontextprotocol/server. A fixed version is available. View diagnostics → Fail
No install/post-install scripts declared.Pass
3 of 46 dependencies flagged as unhealthy. View diagnostics → Partial

Provenance & Transparency100

Source repository is publicly reachable at the declared URL. View diagnostics → Pass
Cryptographically verified build provenance (signed, bound to razzant/claudexor). View diagnostics → Pass
Clear OSI-approved license (MIT).Pass
Actively maintained (last published 1 days ago).Pass
Publishes a security disclosure policy (SECURITY.md).Pass

Schema Quality & AI Usability58

AI-judged instruction clarity (good).Pass
Context-footprint check failed: tool/resource definitions use about 8811 tokens (~214/item across 41 items; 41 tools + 0 resources), over budget; trim descriptions and params. See how to fix → Fail
Usage-examples check failed: none of the tools include examples. See how to fix → Fail

Stability & Change Management100

No destabilizing schema changes in the last 30 days.Pass

Tool Coverage93

100% of tools have a non-trivial description (not blank, and not just the tool's name).Pass
75% of tool parameters carry a description.Partial
Structured output schemas are declared (73% of tools); any adoption earns full credit.Pass

Tool Safety100

No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.Pass
All 1 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation.Pass
An AI judge read all 41 captured unit(s) of tool text and found none that tries to manipulate the model reading it.Pass

Capabilities100

Implements a current MCP spec version (2026-07-28).Pass

Changelog

Every change we have recorded for this component, newest first. Security-relevant changes are always shown. ▲ marks a change for the better, ▼ a change for the worse; unmarked changes are neutral.

4 Nov 26 0
- CVE-2026-111771 affects this package: high ▼ security
- CVE-2026-109924 affects this package: high ▼ security
- CVE-2026-9697 affects this package: high ▼ security
- CVE-2026-109829 affects this package: high ▼ security
- CVE-2026-84933 affects this package: high ▼ security
- CVE-2026-11525 affects this package: high ▼ security
- CVE-2026-13697 affects this package: high ▼ security
- CVE-2026-18149 affects this package: high ▼ security
- CVE-2026-111770 affects this package: high ▼ security
- CVE-2026-109383 affects this package: high ▼ security
- CVE-2026-9679 affects this package: high ▼ security
- CVE-2026-14643 affects this package: high ▼ security
- CVE-2026-16728 affects this package: high ▼ security
- CVE-2026-9678 affects this package: high ▼ security
- CVE-2026-84947 affects this package: high ▼ security
- CVE-2026-84890 affects this package: high ▼ security
- CVE-2026-15157 affects this package: high ▼ security
- CVE-2026-6733 affects this package: high ▼ security
- CVE-2026-16729 affects this package: high ▼ security
- CVE-2026-109869 affects this package: high ▼ security
- CVE-2026-111768 affects this package: high ▼ security
- CVE-2026-85014 affects this package: high ▼ security
- CVE-2026-12151 affects this package: high ▼ security
- CVE-2026-85008 affects this package: high ▼ security
- CVE-2026-18540 affects this package: high ▼ security
- CVE-2026-6734 affects this package: high ▼ security
- CVE-2026-87029 affects this package: high ▼ security
- CVE-2026-109841 affects this package: high ▼ security
- CVE-2026-87018 affects this package: high ▼ security
- CVE-2026-19534 affects this package: high ▼ security
- CVE-2026-84961 affects this package: high ▼ security
- CVE-2026-109932 affects this package: high ▼ security
3 Nov 26 0
- CVE-2026-111771 affects this package: high ▼ security
- CVE-2026-109924 affects this package: high ▼ security
- CVE-2026-9697 affects this package: high ▼ security
- CVE-2026-109829 affects this package: high ▼ security
- CVE-2026-84933 affects this package: high ▼ security
- CVE-2026-11525 affects this package: high ▼ security
- CVE-2026-13697 affects this package: high ▼ security
- CVE-2026-18149 affects this package: high ▼ security
- CVE-2026-111770 affects this package: high ▼ security
- CVE-2026-109383 affects this package: high ▼ security
- CVE-2026-9679 affects this package: high ▼ security
- CVE-2026-14643 affects this package: high ▼ security
- CVE-2026-16728 affects this package: high ▼ security
- CVE-2026-9678 affects this package: high ▼ security
- CVE-2026-84947 affects this package: high ▼ security
- CVE-2026-84890 affects this package: high ▼ security
- CVE-2026-15157 affects this package: high ▼ security
- CVE-2026-6733 affects this package: high ▼ security
- CVE-2026-16729 affects this package: high ▼ security
- CVE-2026-109869 affects this package: high ▼ security
- CVE-2026-111768 affects this package: high ▼ security
- CVE-2026-85014 affects this package: high ▼ security
- CVE-2026-12151 affects this package: high ▼ security
- CVE-2026-85008 affects this package: high ▼ security
- CVE-2026-18540 affects this package: high ▼ security
- CVE-2026-6734 affects this package: high ▼ security
- CVE-2026-87029 affects this package: high ▼ security
- CVE-2026-109841 affects this package: high ▼ security
- CVE-2026-87018 affects this package: high ▼ security
- CVE-2026-19534 affects this package: high ▼ security
- CVE-2026-84961 affects this package: high ▼ security
- CVE-2026-109932 affects this package: high ▼ security
2 Nov 26 0
- CVE-2026-111771 affects this package: high ▼ security
- CVE-2026-109924 affects this package: high ▼ security
- CVE-2026-9697 affects this package: high ▼ security
- CVE-2026-109829 affects this package: high ▼ security
- CVE-2026-84933 affects this package: high ▼ security
- CVE-2026-11525 affects this package: high ▼ security
- CVE-2026-13697 affects this package: high ▼ security
- CVE-2026-18149 affects this package: high ▼ security
- CVE-2026-111770 affects this package: high ▼ security
- CVE-2026-109383 affects this package: high ▼ security
- CVE-2026-9679 affects this package: high ▼ security
- CVE-2026-14643 affects this package: high ▼ security
- CVE-2026-16728 affects this package: high ▼ security
- CVE-2026-9678 affects this package: high ▼ security
- CVE-2026-84947 affects this package: high ▼ security
- CVE-2026-84890 affects this package: high ▼ security
- CVE-2026-15157 affects this package: high ▼ security
- CVE-2026-6733 affects this package: high ▼ security
- CVE-2026-16729 affects this package: high ▼ security
- CVE-2026-109869 affects this package: high ▼ security
- CVE-2026-111768 affects this package: high ▼ security
- CVE-2026-85014 affects this package: high ▼ security
- CVE-2026-12151 affects this package: high ▼ security
- CVE-2026-85008 affects this package: high ▼ security
- CVE-2026-18540 affects this package: high ▼ security
- CVE-2026-6734 affects this package: high ▼ security
- CVE-2026-87029 affects this package: high ▼ security
- CVE-2026-109841 affects this package: high ▼ security
- CVE-2026-87018 affects this package: high ▼ security
- CVE-2026-19534 affects this package: high ▼ security
- CVE-2026-84961 affects this package: high ▼ security
- CVE-2026-109932 affects this package: high ▼ security
1 Nov 26 0
- CVE-2026-111771 affects this package: high ▼ security
- CVE-2026-109924 affects this package: high ▼ security
- CVE-2026-9697 affects this package: high ▼ security
- CVE-2026-109829 affects this package: high ▼ security
- CVE-2026-84933 affects this package: high ▼ security
- CVE-2026-11525 affects this package: high ▼ security
- CVE-2026-13697 affects this package: high ▼ security
- CVE-2026-18149 affects this package: high ▼ security
- CVE-2026-111770 affects this package: high ▼ security
- CVE-2026-109383 affects this package: high ▼ security
- CVE-2026-9679 affects this package: high ▼ security
- CVE-2026-14643 affects this package: high ▼ security
- CVE-2026-16728 affects this package: high ▼ security
- CVE-2026-9678 affects this package: high ▼ security
- CVE-2026-84947 affects this package: high ▼ security
- CVE-2026-84890 affects this package: high ▼ security
- CVE-2026-15157 affects this package: high ▼ security
- CVE-2026-6733 affects this package: high ▼ security
- CVE-2026-16729 affects this package: high ▼ security
- CVE-2026-109869 affects this package: high ▼ security
- CVE-2026-111768 affects this package: high ▼ security
- CVE-2026-85014 affects this package: high ▼ security
- CVE-2026-12151 affects this package: high ▼ security
- CVE-2026-85008 affects this package: high ▼ security
- CVE-2026-18540 affects this package: high ▼ security
- CVE-2026-6734 affects this package: high ▼ security
- CVE-2026-87029 affects this package: high ▼ security
- CVE-2026-109841 affects this package: high ▼ security
- CVE-2026-87018 affects this package: high ▼ security
- CVE-2026-19534 affects this package: high ▼ security
- CVE-2026-84961 affects this package: high ▼ security
- CVE-2026-109932 affects this package: high ▼ security
31 Oct 26 0
- CVE-2026-111771 affects this package: high ▼ security
- CVE-2026-109924 affects this package: high ▼ security
- CVE-2026-9697 affects this package: high ▼ security
- CVE-2026-109829 affects this package: high ▼ security
- CVE-2026-84933 affects this package: high ▼ security
- CVE-2026-11525 affects this package: high ▼ security
- CVE-2026-13697 affects this package: high ▼ security
- CVE-2026-18149 affects this package: high ▼ security
- CVE-2026-111770 affects this package: high ▼ security
- CVE-2026-109383 affects this package: high ▼ security
- CVE-2026-9679 affects this package: high ▼ security
- CVE-2026-14643 affects this package: high ▼ security
- CVE-2026-16728 affects this package: high ▼ security
- CVE-2026-9678 affects this package: high ▼ security
- CVE-2026-84947 affects this package: high ▼ security
- CVE-2026-84890 affects this package: high ▼ security
- CVE-2026-15157 affects this package: high ▼ security
- CVE-2026-6733 affects this package: high ▼ security
- CVE-2026-16729 affects this package: high ▼ security
- CVE-2026-109869 affects this package: high ▼ security
- CVE-2026-111768 affects this package: high ▼ security
- CVE-2026-85014 affects this package: high ▼ security
- CVE-2026-12151 affects this package: high ▼ security
- CVE-2026-85008 affects this package: high ▼ security
- CVE-2026-18540 affects this package: high ▼ security
- CVE-2026-6734 affects this package: high ▼ security
- CVE-2026-87029 affects this package: high ▼ security
- CVE-2026-109841 affects this package: high ▼ security
- CVE-2026-87018 affects this package: high ▼ security
- CVE-2026-19534 affects this package: high ▼ security
- CVE-2026-84961 affects this package: high ▼ security
- CVE-2026-109932 affects this package: high ▼ security
30 Oct 26 0
- CVE-2026-111771 affects this package: high ▼ security
- CVE-2026-109924 affects this package: high ▼ security
- CVE-2026-9697 affects this package: high ▼ security
- CVE-2026-109829 affects this package: high ▼ security
- CVE-2026-84933 affects this package: high ▼ security
- CVE-2026-11525 affects this package: high ▼ security
- CVE-2026-13697 affects this package: high ▼ security
- CVE-2026-18149 affects this package: high ▼ security
- CVE-2026-111770 affects this package: high ▼ security
- CVE-2026-109383 affects this package: high ▼ security
- CVE-2026-9679 affects this package: high ▼ security
- CVE-2026-14643 affects this package: high ▼ security
- CVE-2026-16728 affects this package: high ▼ security
- CVE-2026-9678 affects this package: high ▼ security
- CVE-2026-84947 affects this package: high ▼ security
- CVE-2026-84890 affects this package: high ▼ security
- CVE-2026-15157 affects this package: high ▼ security
- CVE-2026-6733 affects this package: high ▼ security
- CVE-2026-16729 affects this package: high ▼ security
- CVE-2026-109869 affects this package: high ▼ security
- CVE-2026-111768 affects this package: high ▼ security
- CVE-2026-85014 affects this package: high ▼ security
- CVE-2026-12151 affects this package: high ▼ security
- CVE-2026-85008 affects this package: high ▼ security
- CVE-2026-18540 affects this package: high ▼ security
- CVE-2026-6734 affects this package: high ▼ security
- CVE-2026-87029 affects this package: high ▼ security
- CVE-2026-109841 affects this package: high ▼ security
- CVE-2026-87018 affects this package: high ▼ security
- CVE-2026-19534 affects this package: high ▼ security
- CVE-2026-84961 affects this package: high ▼ security
- CVE-2026-109932 affects this package: high ▼ security
- Stability: 0.97 → pass security
29 Oct 26 0

We did not load change detail this far back for this component, so this day may have recorded more than is shown.
- CVE-2026-111771 affects this package: high ▼ security
- CVE-2026-109924 affects this package: high ▼ security
- CVE-2026-9697 affects this package: high ▼ security
- CVE-2026-109829 affects this package: high ▼ security
- CVE-2026-84933 affects this package: high ▼ security
- CVE-2026-11525 affects this package: high ▼ security
- CVE-2026-13697 affects this package: high ▼ security
28 Oct 26 −1

We did not load change detail this far back for this component, so this day may have recorded more than is shown.

Show 1 earlier entry Hide earlier entries

Diagnostics

Diagnostic detail from the automated scan of this channel: what the scanner observed at each step, so you can see exactly where a check passed or failed. It is informational only and never changes the trust score.

Captured 4 Nov 2026 · Analysed npm/claudexor@4.8.1

Provenance Verified

A signed build attestation was found and verified, binding this exact artifact to the source repository it claims to come from.

Result	Verified
Ecosystem	npm
Reason	Verified
Discovered via	Registry attestation endpoint
Source repo	razzant/claudexor
Certificate issuer	https://token.actions.githubusercontent.com
Certificate SAN	https://github.com/razzant/claudexor/.github/workflows/release.yml@refs/tags/v4.8.1
Rekor log index	3535796329
Predicate type	SLSA build provenance https://slsa.dev/provenance/v1
Subject digest	sha512:e84c3341b79fded72df7d2180f24cd7028e7f67b39d198cae85dad10907312658da3019c789ad9f645170cf9eda66db86e5486391b2e0a5e611a4ddfd

Background: How many MCP packages publish verified provenance →

Vulnerabilities 32 findings

ID	CVE	Severity	Vector	Fix available
GHSA-46jr-4wx8-v453	CVE-2026-111768	high	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-jm63-cq47-j5jp	CVE-2026-111770	high	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:L/A:H	yes
GHSA-rvq5-wwqv-78pq	CVE-2026-111771	medium	CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:N	yes
GHSA-2gqq-gqf2-x968	CVE-2026-84947	low	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N	yes
GHSA-2jfj-6hjv-fm6j	CVE-2026-84933	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:L/A:N	yes
GHSA-2m5v-pch6-74hx	CVE-2026-109383	high	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:N	yes
GHSA-35p6-xmwp-9g52	CVE-2026-6733	low	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N	yes
GHSA-3xpg-4rpp-hhhm	CVE-2026-84890	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-4cwx-7wf7-3272	CVE-2026-13697	high	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:H	yes
GHSA-4qv8-2p9j-2685	CVE-2026-109924	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-4w7w-5xrc-9h26	CVE-2026-109829	high	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:N	yes
GHSA-55wc-3pg8-jfw6	CVE-2026-87029	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-8436-99hf-9mmv	CVE-2026-85008	low	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N	yes
GHSA-8xcm-r25x-g524	CVE-2026-16728	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N	yes
GHSA-9fm3-97px-rjf6	CVE-2026-109932	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-f72v-6w45-q472	CVE-2026-109841	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:N	yes
GHSA-g8m3-5g58-fq7m	CVE-2026-11525	low	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N	yes
GHSA-hm92-r4w5-c3mj	CVE-2026-6734	high	CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H	yes
GHSA-jr45-8vmc-qm54	CVE-2026-14643	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N	yes
GHSA-m8rv-5g2x-5cg5	CVE-2026-15157	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:L/I:L/A:N	yes
GHSA-p88m-4jfj-68fv	CVE-2026-9679	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:N	yes
GHSA-pmjh-fq2x-6v4x	CVE-2026-18149	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-pr7r-676h-xcf6	CVE-2026-9678	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N	yes
GHSA-qrxm-fqhp-jwvm	CVE-2026-87018	high	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-r53p-7pc4-xj5r	CVE-2026-18540	low	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N	yes
GHSA-rfgv-xxqx-mfg5	CVE-2026-19534	high	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-rx4f-c7p8-82vq	CVE-2026-85014	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-v3r7-h72x-cjcm	CVE-2026-16729	medium	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N	yes
GHSA-vmh5-mc38-953g	CVE-2026-9697	high	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:N	yes
GHSA-vxpw-j846-p89q	CVE-2026-12151	high	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H	yes
GHSA-w293-vg96-wgc3	CVE-2026-84961	high	CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:N	yes
GHSA-xjj6-f33q-929c	CVE-2026-109869	high	CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H	yes

Background: What a vulnerability scan can and cannot prove →

Dependencies 46 packages

Packages resolved	46
Stale	3
Tree resolution	Complete

Background: SBOMs and build attestations, explained →

MCP tools · 41 exposed · ~8,811 tokens

The tools this component advertises to a client, with an estimated token cost for each. Expand a tool to see its parameters and schema. The per-tool counts are indicative and are not scored directly; the schema's total context footprint is one signal in Schema Quality & AI Usability. A tool's description is untrusted text the model reads on every call, which is what makes this list a security surface and not just an inventory: how tool poisoning works →

Tool Tokens

claudexor_account_reset ~132

Explicitly consume a selected account reset, outside inference capacity. Supply target, offer_id, optional grant_id and idempotency_key. Retain the original key and request after a timeout and repeat them to recover; never replace the key automatically. Or supply operation_id to read a receipt. Provider outcome and resource readback are separate; Claude already_used is not proof of this operation's success.

Input Output Examples

Name	Type	Req	Description
grant_id	string	–	–
idempotency_key	string	–	–
offer_id	string	–	–
operation_id	string	–	–
target	object	–	–

Name	Type	Req	Description
completed_at	–	yes	–
created_at	string	yes	–
detail	string\|null	yes	–
id	string	yes	–
outcome	string	yes	–
readback	object	yes	–
request	object	yes	–
resources	–	yes	Current quota snapshots without a fabricated aggregate.
state	string	yes	–

No examples provided.

claudexor_account_resources ~71

Read typed account balances, spending, reset offers and quota. refresh:true requests provider data; target selects one exact registered account, including a disabled portable profile. No login or spending-policy change.

Input Output Examples

Name	Type	Req	Description
model	string	–	–
refresh	boolean	–	–
target	object	–	–

Name	Type	Req	Description
absences	array	–	Every registered subject reports a snapshot or a typed absence; a refresh gap may coexist with a stale last-known snapshot.
refresh_skipped	array	–	–
refreshed_at	–	yes	–
resources	array	yes	–
snapshots	array	yes	–

No examples provided.

claudexor_accounts ~260

Return the read-only Accounts view: registered profiles, readiness, quota freshness, and routing identity. accountPools.next_up is the configured default model's verdict, never a forecast for another model; routingEvidence carries model-scoped recent refusals, temporary per-model capability marks, and each account's last service per requested model (a routing observation, not a quota or authentication verdict). Default combines current quota and registry facts with retained readiness observations: one coalesced cold acquisition, including failure, then passive reads with the original check time. Viewing does not renew old evidence; actual execution performs its own checks. fresh:true requests the atomic snapshot instead — an EXPENSIVE explicit refresh (a live probe per registered profile, a full harness doctor sweep, and the vendor quota fan-out, which honors account/route rate-limit cooldowns and existing broad floors and discloses paused subjects in quota.refresh_skipped). Use fresh:true only when acting on staleness matters; never poll with it. This tool never starts login or changes account state.

Input Output Examples

Name	Type	Req	Description
fresh	boolean	–	true = the expensive atomic Accounts snapshot (fresh probes + doctor + vendor quota fan-out); false/absent = current quota/registry facts with retained readiness observations and their age.

Structured output declared, but exposes no named fields.

No examples provided.

claudexor_answer_interaction ~102

Answer a daemon-persisted run interaction; success (status delivered) is reported only after the control API acknowledges the journal mutation. Refusals keep their typed status as the failure code: not_found, already_resolved (never proof that this answer landed) or rejected. Answers have no replay key and are not retried.

Input Output Examples

Name	Type	Req	Description
answers	array	yes	–
interactionId	string	yes	–
runId	string	yes	–

No output schema declared.

No examples provided.

claudexor_apply_check ~58

Dry-check whether a run's patch would apply cleanly to its original project (no mutation).

Input Output Examples

Name	Type	Req	Description
runId	string	yes	Daemon run id (from a run tool's runId trailer or claudexor_runs).

No output schema declared.

No examples provided.

claudexor_ask ~576

Enqueue a read-only Claudexor answer; pass deepScan:true for a bounded multi-scout research sweep with synthesis. Returns a durable run handle, not terminal output; use claudexor_run_status/result for the final answer.

Input Output Examples

Name	Type	Req	Description
access	string	–	Optional access profile for this run.
attachments	array	–	Immutable resource references from upload finalize. Paths, base64 and internal resolved attachments are rejected; ordinary lane admission still applies.
credentialProfileId	string	–	Strict credential profile for this run; never falls back to another account.
deepScan	boolean	–	Widen the answer into a bounded multi-scout research sweep with synthesis.
effort	string	–	Optional effort override for the primary harness. Expects a reasoning-effort level the resolved harness/model advertises (a lowercase slug; each vendor lists its own levels, weakest to strongest); an…
execution	object	–	–
externalContextPolicy	string	–	Alias of web for control-api parity.
harness	string	–	Optional harness id to force for this one-shot run.
idempotencyKey	string	–	Caller-owned key for this exact launch; requires an explicit repoPath. After an unknown response, repeat the same key and arguments to recover the accepted job or run (a queued job returns its jobId)…
maxSeconds	integer	–	Initial run time limit in seconds. Explicit limits default to hard unless timeLimitMode is soft.
model	string	–	Optional model override for the primary harness.
n	integer	–	Optional best-of-N width for candidate races (or deep-scan scout width).
paidBudget	–	–	Incremental-cash budget: explicitly unlimited or a finite non-negative USD cap.
primaryHarness	string	–	Optional primary harness id for this run.
processingPreference	string	–	Advisory service preference; omission inherits captured settings and explicit Standard requests ordinary service.
prompt	string	yes	The user task or question to run through Claudexor.
repoPath	string	–	Absolute path of the target project. Defaults to the MCP server cwd; required with idempotencyKey.
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.
timeLimitMode	string	–	With maxSeconds: soft warns and continues; hard cancels on expiry.
web	string	–	External context policy for this run.

Name	Type	Req	Description
applyEligibility	–	yes	Apply-gate verdict for mutating runs; null for read-only routes or when no patch exists.
attemptExecution	array	–	–
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan tools or deferred handles.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	–
delegation	–	–	Engine-owned Delegate requested/effective/used receipt; surfaces project it without parsing model output.
detailProblem	–	–	Typed post-terminal detail-read problem (the run result survives; its detail projections are absent for this reason); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
jobId	string\|null	–	Daemon job id of the accepted launch. While runId is null, pass it as runId to the status, wait and result tools; they report the canonical runId once the run binds.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); the single honest one-line verdict, null while non-terminal or unavailable.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges); null for non-terminal or read-only routes.
parentRunId	string\|null	–	–
planReadiness	–	–	Derived plan readiness for plan tools (D17); null for non-plan tools.
runDir	string\|null	yes	Artifact directory of the run; null when not persisted.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	yes	Daemon run id once the accepted job started a run (recovery handle for inspect/follow/apply/decision); null while the job is still queued or after it failed before a run started.
status	string\|null	yes	Run lifecycle state (queued \| running \| succeeded \| failed \| cancelled \| interrupted); null when an accepted job's state could not be read.
storeId	string	–	Identity of the durable engine store that accepted this group member, stable across restarts.
summary	string	yes	Human-readable outcome summary (same text as the tool's text content).
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.

No examples provided.

claudexor_best_of ~672

Enqueue a best-of-N Claudexor run with cross-family review. Returns a durable run handle; use claudexor_run_status/result for terminal output.

Input Output Examples

Name	Type	Req	Description
access	string	–	Optional access profile for this run.
attachments	array	–	Immutable resource references from upload finalize. Paths, base64 and internal resolved attachments are rejected; ordinary lane admission still applies.
credentialProfileId	string	–	Strict credential profile for this run; never falls back to another account.
effort	string	–	Optional effort override for the primary harness. Expects a reasoning-effort level the resolved harness/model advertises (a lowercase slug; each vendor lists its own levels, weakest to strongest); an…
execution	object	–	–
externalContextPolicy	string	–	Alias of web for control-api parity.
harness	string	–	Optional harness id to force for this one-shot run.
idempotencyKey	string	–	Caller-owned key for this exact launch; requires an explicit repoPath. After an unknown response, repeat the same key and arguments to recover the accepted job or run (a queued job returns its jobId)…
maxSeconds	integer	–	Initial run time limit in seconds. Explicit limits default to hard unless timeLimitMode is soft.
model	string	–	Optional model override for the primary harness.
n	integer	–	Optional best-of-N width for candidate races (or deep-scan scout width).
paidBudget	–	–	Incremental-cash budget: explicitly unlimited or a finite non-negative USD cap.
primaryHarness	string	–	Optional primary harness id for this run.
processingPreference	string	–	Advisory service preference; omission inherits captured settings and explicit Standard requests ordinary service.
prompt	string	yes	The user task or question to run through Claudexor.
protectedPathApprovals	array	–	Typed approvals for existing protected path edits.
repoPath	string	–	Absolute path of the target project. Defaults to the MCP server cwd; required with idempotencyKey.
review	boolean	–	Enable internal model review. Ordinary Agent defaults to false; true selects reviewers automatically. An explicit reviewer panel or reviewer overrides also enable review. Best-of and until-clean reta…
reviewerEfforts	object	–	Enable review with per-provider reviewer effort overrides.
reviewerModels	object	–	Enable review with per-provider reviewer model overrides.
reviewerPanel	array	–	Enable review with explicit panel entries, preserving order and duplicates; no additional review flag is needed.
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.
tests	array	–	Typed-argv deterministic gate commands for this run.
timeLimitMode	string	–	With maxSeconds: soft warns and continues; hard cancels on expiry.
web	string	–	External context policy for this run.

Name	Type	Req	Description
applyEligibility	–	yes	Apply-gate verdict for mutating runs; null for read-only routes or when no patch exists.
attemptExecution	array	–	–
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan tools or deferred handles.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	–
delegation	–	–	Engine-owned Delegate requested/effective/used receipt; surfaces project it without parsing model output.
detailProblem	–	–	Typed post-terminal detail-read problem (the run result survives; its detail projections are absent for this reason); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
jobId	string\|null	–	Daemon job id of the accepted launch. While runId is null, pass it as runId to the status, wait and result tools; they report the canonical runId once the run binds.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); the single honest one-line verdict, null while non-terminal or unavailable.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges); null for non-terminal or read-only routes.
parentRunId	string\|null	–	–
planReadiness	–	–	Derived plan readiness for plan tools (D17); null for non-plan tools.
runDir	string\|null	yes	Artifact directory of the run; null when not persisted.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	yes	Daemon run id once the accepted job started a run (recovery handle for inspect/follow/apply/decision); null while the job is still queued or after it failed before a run started.
status	string\|null	yes	Run lifecycle state (queued \| running \| succeeded \| failed \| cancelled \| interrupted); null when an accepted job's state could not be read.
storeId	string	–	Identity of the durable engine store that accepted this group member, stable across restarts.
summary	string	yes	Human-readable outcome summary (same text as the tool's text content).
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.

No examples provided.

claudexor_capabilities ~51

Return the derived AgentCapabilityCatalog: per-harness live capabilities (doctor-backed), canonical modes, the mutability matrix, run-control keys, CLI verbs, and the run-apply-state vocabulary.

Input Output Examples

Input schema present but exposes no named parameters.

Name	Type	Req	Description
availableHarnesses	array	yes	Convenience: ids with doctor status ok.
cliCommands	array	yes	CLI verbs with mutability/stability (full flag detail via `claudexor help --json`).
generatedAt	string	yes	ISO timestamp when the catalog was composed.
git	object	yes	Execution-location Git readiness. Read-only and valid non-Git in-place flows do not require it.
harnesses	array	yes	Live per-harness capabilities (doctor-backed).
mcpTools	array	yes	MCP tool names `claudexor mcp serve` exposes.
modes	array	yes	Canonical run modes; strategy controls are separate schema-owned fields, never modes.
mutability	object	yes	The mutability matrix: every way a Claudexor run can (or cannot) touch a tree, from the schema's closed vocabularies.
ok	boolean	yes	Envelope marker (matches the CLI JSON convention).
outputSchemaDialects	array	yes	Supported structured-output JSON Schema dialects and canonical $schema URIs.
runApplyStates	array	yes	RunApplyState vocabulary an agent can encounter on run results (not_applied \| applied \| applied_review_blocked \| reverted) — distinct from the ApplyEligibility verdict object on run details.
runControlKeys	array	yes	Accepted POST /runs request keys (derived from the ControlRunStartRequest schema; the CLI/MCP/ACP surfaces project subsets of these).
version	string	yes	Claudexor version serving this catalog (compare with CLAUDEXOR_PLUGIN_VERSION to detect plugin skew).

No examples provided.

claudexor_create ~668

Enqueue a create-from-scratch Claudexor run. Returns a durable run handle; use claudexor_run_status/result for terminal output.

Input Output Examples

Name	Type	Req	Description
access	string	–	Optional access profile for this run.
attachments	array	–	Immutable resource references from upload finalize. Paths, base64 and internal resolved attachments are rejected; ordinary lane admission still applies.
credentialProfileId	string	–	Strict credential profile for this run; never falls back to another account.
effort	string	–	Optional effort override for the primary harness. Expects a reasoning-effort level the resolved harness/model advertises (a lowercase slug; each vendor lists its own levels, weakest to strongest); an…
execution	object	–	–
externalContextPolicy	string	–	Alias of web for control-api parity.
harness	string	–	Optional harness id to force for this one-shot run.
idempotencyKey	string	–	Caller-owned key for this exact launch; requires an explicit repoPath. After an unknown response, repeat the same key and arguments to recover the accepted job or run (a queued job returns its jobId)…
maxSeconds	integer	–	Initial run time limit in seconds. Explicit limits default to hard unless timeLimitMode is soft.
model	string	–	Optional model override for the primary harness.
n	integer	–	Optional best-of-N width for candidate races (or deep-scan scout width).
paidBudget	–	–	Incremental-cash budget: explicitly unlimited or a finite non-negative USD cap.
primaryHarness	string	–	Optional primary harness id for this run.
processingPreference	string	–	Advisory service preference; omission inherits captured settings and explicit Standard requests ordinary service.
prompt	string	yes	The user task or question to run through Claudexor.
protectedPathApprovals	array	–	Typed approvals for existing protected path edits.
repoPath	string	–	Absolute path of the target project. Defaults to the MCP server cwd; required with idempotencyKey.
review	boolean	–	Enable internal model review. Ordinary Agent defaults to false; true selects reviewers automatically. An explicit reviewer panel or reviewer overrides also enable review. Best-of and until-clean reta…
reviewerEfforts	object	–	Enable review with per-provider reviewer effort overrides.
reviewerModels	object	–	Enable review with per-provider reviewer model overrides.
reviewerPanel	array	–	Enable review with explicit panel entries, preserving order and duplicates; no additional review flag is needed.
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.
tests	array	–	Typed-argv deterministic gate commands for this run.
timeLimitMode	string	–	With maxSeconds: soft warns and continues; hard cancels on expiry.
web	string	–	External context policy for this run.

Name	Type	Req	Description
applyEligibility	–	yes	Apply-gate verdict for mutating runs; null for read-only routes or when no patch exists.
attemptExecution	array	–	–
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan tools or deferred handles.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	–
delegation	–	–	Engine-owned Delegate requested/effective/used receipt; surfaces project it without parsing model output.
detailProblem	–	–	Typed post-terminal detail-read problem (the run result survives; its detail projections are absent for this reason); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
jobId	string\|null	–	Daemon job id of the accepted launch. While runId is null, pass it as runId to the status, wait and result tools; they report the canonical runId once the run binds.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); the single honest one-line verdict, null while non-terminal or unavailable.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges); null for non-terminal or read-only routes.
parentRunId	string\|null	–	–
planReadiness	–	–	Derived plan readiness for plan tools (D17); null for non-plan tools.
runDir	string\|null	yes	Artifact directory of the run; null when not persisted.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	yes	Daemon run id once the accepted job started a run (recovery handle for inspect/follow/apply/decision); null while the job is still queued or after it failed before a run started.
status	string\|null	yes	Run lifecycle state (queued \| running \| succeeded \| failed \| cancelled \| interrupted); null when an accepted job's state could not be read.
storeId	string	–	Identity of the durable engine store that accepted this group member, stable across restarts.
summary	string	yes	Human-readable outcome summary (same text as the tool's text content).
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.

No examples provided.

claudexor_inspect ~64

Inspect a daemon-tracked run: status, summary, decision verdict, and the derived applyEligibility (what unblocks apply).

Input Output Examples

Name	Type	Req	Description
runId	string	yes	Daemon run id (from a run tool's runId trailer or claudexor_runs).

Name	Type	Req	Description
applyEligibility	–	–	Derived apply-gate verdict; null for read-only runs or when no patch exists.
attemptExecution	array	–	–
budget	–	–	Run budget snapshot (cash + subscription valuation, QA-023c); null when unavailable.
continueFrom	–	–	Predecessor run this run continues; null for an ordinary run, omitted when the detail did not project it.
continuity	array	–	Per-try continuation receipts (carrier, cause, accounts, model) from the run's event log; omitted when not projected.
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan runs.
decisionStatus	string\|null	–	The arbitration decision status, when the run reached arbitration.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	Claudexor Delegate parent id; null for ordinary runs and native subagents.
delegation	–	–	Typed Delegate receipt; null when unavailable or legacy.
detailProblem	–	–	Typed post-terminal detail-read problem (the durable run handle survives); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); null while non-terminal.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges), when terminal.
parentRunId	string\|null	–	Server-owned ordinary parent/follow-up run id, when present.
pendingInteractions	–	–	Count of questions still awaiting answers, when known.
planReadiness	–	–	Derived plan readiness (plan runs only, D17); null otherwise.
resumable	–	–	Terminal facts of stopped unfinished work that claudexor_run_continue can continue (cause, resetsAt, carriers, session, workspace); null when the work finished or the run is not terminal, omitted whe…
retainedEnvelope	–	–	Isolated envelope kept for a successor (disk use visible); null when none, omitted when not projected.
runDir	string\|null	–	Artifact directory; null when absent.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	–	The daemon run id, when known.
secretLike	–	–	INV-062 disclosure: secret-like strings kept in the work but hidden in saved copies (paths and counts only); null when nothing was hidden.
status	string\|null	–	The run's terminal or in-flight lifecycle state, when known.
summary	string	yes	Human-readable outcome/status text (same as the tool's text content).

No examples provided.

claudexor_journal_recovery ~42

Inspect, validate, or export one durable journal partition through v2.

Input Output Examples

Name	Type	Req	Description
action	string	yes	–
partition	string	yes	–

No output schema declared.

No examples provided.

claudexor_next_message ~198

Operator-only editable next message for one daemon-owned single-candidate Agent/Ask predecessor. request.action create captures controls once and reserves its one successor; replace changes only the full prompt/ref list at expectedRevision; clear releases only the draft. Every mutation requires a stable idempotencyKey. Eligible settled success dispatches one ordinary continuation; failures, Stop, restart or admission refusal hold without retry. Held edits never rearm. The successor keeps the predecessor's execution: after a successful isolated (envelope) Agent it starts from the project base and receives only the path of the unapplied saved patch; nothing is applied automatically. No thread or engine Delegate child targets. Ordinary controller-owned one-shot workers are supported.

Input Output Examples

Name	Type	Req	Description
idempotencyKey	string	yes	–
request	–	yes	One editable reserved next message of a daemon-owned one-shot Agent/Ask. Every mutation requires Idempotency-Key; replace/clear require the exact revision.

Name	Type	Req	Description
appliedRevision	integer	yes	The exact revision originally applied by this idempotency key.
draft	object	yes	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.

No examples provided.

claudexor_next_message_read ~109

Read one exact deferred draft: bounded prompt/ref preview with explicit omissions, predecessor result path, successor handles and `controls`, the values frozen at creation by origin (inherited from the predecessor, resolved at creation, captured Settings preferences) with their capture time. Permissions, trust, accounts and quota are not frozen. This read never dispatches. A terminal predecessor may still have a waiting or held draft; inspect its state.

Input Output Examples

Name	Type	Req	Description
draftId	string	yes	Non-empty identifier string.

Name	Type	Req	Description
attachments	array	yes	–
controls	–	yes	Null after clear.
detailPath	string	yes	–
draft	object	yes	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
omittedAttachments	integer	yes	–
omittedPromptBytes	integer	yes	–
promptBytes	integer	yes	–
promptPreview	string	yes	–

No examples provided.

claudexor_plan ~559

Enqueue a read-only Claudexor implementation plan. Returns a durable run handle; use claudexor_run_status/result for the final plan.

Input Output Examples

Name	Type	Req	Description
access	string	–	Optional access profile for this run.
attachments	array	–	Immutable resource references from upload finalize. Paths, base64 and internal resolved attachments are rejected; ordinary lane admission still applies.
council	boolean	–	Draft n plans (2 to the configured cap), then merge one plan and question set.
credentialProfileId	string	–	Strict credential profile for this run; never falls back to another account.
effort	string	–	Optional effort override for the primary harness. Expects a reasoning-effort level the resolved harness/model advertises (a lowercase slug; each vendor lists its own levels, weakest to strongest); an…
execution	object	–	–
externalContextPolicy	string	–	Alias of web for control-api parity.
harness	string	–	Optional harness id to force for this one-shot run.
idempotencyKey	string	–	Caller-owned key for this exact launch; requires an explicit repoPath. After an unknown response, repeat the same key and arguments to recover the accepted job or run (a queued job returns its jobId)…
maxSeconds	integer	–	Initial run time limit in seconds. Explicit limits default to hard unless timeLimitMode is soft.
model	string	–	Optional model override for the primary harness.
n	integer	–	Optional best-of-N width for candidate races (or deep-scan scout width).
paidBudget	–	–	Incremental-cash budget: explicitly unlimited or a finite non-negative USD cap.
primaryHarness	string	–	Optional primary harness id for this run.
processingPreference	string	–	Advisory service preference; omission inherits captured settings and explicit Standard requests ordinary service.
prompt	string	yes	The user task or question to run through Claudexor.
repoPath	string	–	Absolute path of the target project. Defaults to the MCP server cwd; required with idempotencyKey.
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.
timeLimitMode	string	–	With maxSeconds: soft warns and continues; hard cancels on expiry.
web	string	–	External context policy for this run.

Name	Type	Req	Description
applyEligibility	–	yes	Apply-gate verdict for mutating runs; null for read-only routes or when no patch exists.
attemptExecution	array	–	–
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan tools or deferred handles.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	–
delegation	–	–	Engine-owned Delegate requested/effective/used receipt; surfaces project it without parsing model output.
detailProblem	–	–	Typed post-terminal detail-read problem (the run result survives; its detail projections are absent for this reason); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
jobId	string\|null	–	Daemon job id of the accepted launch. While runId is null, pass it as runId to the status, wait and result tools; they report the canonical runId once the run binds.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); the single honest one-line verdict, null while non-terminal or unavailable.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges); null for non-terminal or read-only routes.
parentRunId	string\|null	–	–
planReadiness	–	–	Derived plan readiness for plan tools (D17); null for non-plan tools.
runDir	string\|null	yes	Artifact directory of the run; null when not persisted.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	yes	Daemon run id once the accepted job started a run (recovery handle for inspect/follow/apply/decision); null while the job is still queued or after it failed before a run started.
status	string\|null	yes	Run lifecycle state (queued \| running \| succeeded \| failed \| cancelled \| interrupted); null when an accepted job's state could not be read.
storeId	string	–	Identity of the durable engine store that accepted this group member, stable across restarts.
summary	string	yes	Human-readable outcome summary (same text as the tool's text content).
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.

No examples provided.

claudexor_quarantine_journal ~50

Quarantine a corrupt partition and start a fresh epoch after exact confirmation.

Input Output Examples

Name	Type	Req	Description
confirmation	string	yes	–
expectedFingerprint	string	yes	–
partition	string	yes	–

No output schema declared.

No examples provided.

claudexor_run ~696

Enqueue an Agent-mode Claudexor run. Internal model review is off by default; review:true selects a panel automatically, and an explicit reviewerPanel enables review. Returns a durable run handle plus any immediate start facts; use claudexor_run_status/result for the terminal WorkProduct.

Input Output Examples

Name	Type	Req	Description
access	string	–	Optional access profile for this run.
attachments	array	–	Immutable resource references from upload finalize. Paths, base64 and internal resolved attachments are rejected; ordinary lane admission still applies.
credentialProfileId	string	–	Strict credential profile for this run; never falls back to another account.
effort	string	–	Optional effort override for the primary harness. Expects a reasoning-effort level the resolved harness/model advertises (a lowercase slug; each vendor lists its own levels, weakest to strongest); an…
execution	object	–	–
externalContextPolicy	string	–	Alias of web for control-api parity.
harness	string	–	Optional harness id to force for this one-shot run.
idempotencyKey	string	–	Caller-owned key for this exact launch; requires an explicit repoPath. After an unknown response, repeat the same key and arguments to recover the accepted job or run (a queued job returns its jobId)…
maxSeconds	integer	–	Initial run time limit in seconds. Explicit limits default to hard unless timeLimitMode is soft.
model	string	–	Optional model override for the primary harness.
n	integer	–	Optional best-of-N width for candidate races (or deep-scan scout width).
paidBudget	–	–	Incremental-cash budget: explicitly unlimited or a finite non-negative USD cap.
primaryHarness	string	–	Optional primary harness id for this run.
processingPreference	string	–	Advisory service preference; omission inherits captured settings and explicit Standard requests ordinary service.
prompt	string	yes	The user task or question to run through Claudexor.
protectedPathApprovals	array	–	Typed approvals for existing protected path edits.
repoPath	string	–	Absolute path of the target project. Defaults to the MCP server cwd; required with idempotencyKey.
review	boolean	–	Enable internal model review. Ordinary Agent defaults to false; true selects reviewers automatically. An explicit reviewer panel or reviewer overrides also enable review. Best-of and until-clean reta…
reviewerEfforts	object	–	Enable review with per-provider reviewer effort overrides.
reviewerModels	object	–	Enable review with per-provider reviewer model overrides.
reviewerPanel	array	–	Enable review with explicit panel entries, preserving order and duplicates; no additional review flag is needed.
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.
tests	array	–	Typed-argv deterministic gate commands for this run.
timeLimitMode	string	–	With maxSeconds: soft warns and continues; hard cancels on expiry.
web	string	–	External context policy for this run.

Name	Type	Req	Description
applyEligibility	–	yes	Apply-gate verdict for mutating runs; null for read-only routes or when no patch exists.
attemptExecution	array	–	–
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan tools or deferred handles.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	–
delegation	–	–	Engine-owned Delegate requested/effective/used receipt; surfaces project it without parsing model output.
detailProblem	–	–	Typed post-terminal detail-read problem (the run result survives; its detail projections are absent for this reason); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
jobId	string\|null	–	Daemon job id of the accepted launch. While runId is null, pass it as runId to the status, wait and result tools; they report the canonical runId once the run binds.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); the single honest one-line verdict, null while non-terminal or unavailable.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges); null for non-terminal or read-only routes.
parentRunId	string\|null	–	–
planReadiness	–	–	Derived plan readiness for plan tools (D17); null for non-plan tools.
runDir	string\|null	yes	Artifact directory of the run; null when not persisted.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	yes	Daemon run id once the accepted job started a run (recovery handle for inspect/follow/apply/decision); null while the job is still queued or after it failed before a run started.
status	string\|null	yes	Run lifecycle state (queued \| running \| succeeded \| failed \| cancelled \| interrupted); null when an accepted job's state could not be read.
storeId	string	–	Identity of the durable engine store that accepted this group member, stable across restarts.
summary	string	yes	Human-readable outcome summary (same text as the tool's text content).
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.

No examples provided.

claudexor_run_cancel ~147

Request cancellation of a daemon-owned run; success is returned only after the control API acknowledges the durable command. The same call holds the run's next message (claudexor_next_message) that is not accepted yet, also after the run ended; an already accepted successor fails next_message_dispatched with its job handle and keeps running. If that hold cannot be saved, the call fails next_message_hold_failed (retryable) after any cancel was applied; the next message is still not sent, and a retried cancel records the hold.

Input Output Examples

Name	Type	Req	Description
runId	string	yes	Daemon run id (from a run tool's runId trailer or claudexor_runs).

No output schema declared.

No examples provided.

claudexor_run_continue ~570

Start one successor of a terminal single-candidate Agent or Ask run on the existing daemon. The daemon inherits only mode, scope, execution, harness, model and the prompt-cache lifetime, which this tool cannot change (another lifetime, Default included, needs a new run); every other omitted control takes its ordinary new-run default, not the predecessor's value, so restate access, web, account, budget, time limit and review when the successor must keep them. Budget and time limits apply to this successor alone, never as a remainder of the chain. prompt is optional and may be empty. Supply idempotencyKey to replay an unknown response. Returns the same durable handle as the launch tools (jobId, runId once bound); pass either as runId to claudexor_run_wait/status/result.

Input Output Examples

Name	Type	Req	Description
access	string	–	Optional access profile for this run.
attachments	array	–	Immutable resource references from upload finalize. Paths, base64 and internal resolved attachments are rejected; ordinary lane admission still applies.
continueCarrier	string	–	auto: the best carrier the engine can prove (native session, moved session, then the evidence index); packet: a fresh session re-briefed by the evidence index (the caller's choice for a derailed sess…
credentialProfileId	string	–	Strict credential profile for this run; never falls back to another account.
externalContextPolicy	string	–	Alias of web for control-api parity.
idempotencyKey	string	–	Reuse the same key and arguments after an unknown response.
maxSeconds	integer	–	Initial run time limit in seconds. Explicit limits default to hard unless timeLimitMode is soft.
paidBudget	–	–	Incremental-cash budget: explicitly unlimited or a finite non-negative USD cap.
prompt	string	–	Optional continuation text; empty is valid.
review	boolean	–	Enable internal model review. Ordinary Agent defaults to false; true selects reviewers automatically. An explicit reviewer panel or reviewer overrides also enable review. Best-of and until-clean reta…
reviewerEfforts	object	–	Enable review with per-provider reviewer effort overrides.
reviewerModels	object	–	Enable review with per-provider reviewer model overrides.
reviewerPanel	array	–	Enable review with explicit panel entries, preserving order and duplicates; no additional review flag is needed.
runId	string	yes	–
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.
timeLimitMode	string	–	With maxSeconds: soft warns and continues; hard cancels on expiry.
web	string	–	External context policy for this run.

Name	Type	Req	Description
applyEligibility	–	yes	Apply-gate verdict for mutating runs; null for read-only routes or when no patch exists.
attemptExecution	array	–	–
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan tools or deferred handles.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	–
delegation	–	–	Engine-owned Delegate requested/effective/used receipt; surfaces project it without parsing model output.
detailProblem	–	–	Typed post-terminal detail-read problem (the run result survives; its detail projections are absent for this reason); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
jobId	string\|null	–	Daemon job id of the accepted launch. While runId is null, pass it as runId to the status, wait and result tools; they report the canonical runId once the run binds.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); the single honest one-line verdict, null while non-terminal or unavailable.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges); null for non-terminal or read-only routes.
parentRunId	string\|null	–	–
planReadiness	–	–	Derived plan readiness for plan tools (D17); null for non-plan tools.
runDir	string\|null	yes	Artifact directory of the run; null when not persisted.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	yes	Daemon run id once the accepted job started a run (recovery handle for inspect/follow/apply/decision); null while the job is still queued or after it failed before a run started.
status	string\|null	yes	Run lifecycle state (queued \| running \| succeeded \| failed \| cancelled \| interrupted); null when an accepted job's state could not be read.
storeId	string	–	Identity of the durable engine store that accepted this group member, stable across restarts.
summary	string	yes	Human-readable outcome summary (same text as the tool's text content).
taskGroupId	string	–	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.

No examples provided.

claudexor_run_event_read ~220

Read one bounded part of one run event by run id and seq, typically claudexor_run_wait's oversizedEvent. Returns the exact UTF-8 bytes [offset, nextOffset) of the redacted serialization the run event stream delivers, with totalBytes and the sha256 of the whole event; parts never split a character. Read from offset 0, then each nextOffset until it equals totalBytes; a partial read is not a consumed event, so do not move claudexor_run_wait past it until the last part. Only validated events of this run are readable; there is no path, artifact or whole-log access.

Input Output Examples

Name	Type	Req	Description
maxBytes	integer	–	Upper bound of this part's UTF-8 bytes; a part may end earlier at a character boundary.
offset	integer	–	0, or the previous part's nextOffset.
runId	string	yes	–
seq	integer	yes	The event to read, e.g. claudexor_run_wait's oversizedEvent.seq.

Name	Type	Req	Description
nextOffset	integer	yes	Where the next part starts; equal to totalBytes when this part ends the event.
offset	integer	yes	UTF-8 byte offset where this part starts.
runId	string	yes	Canonical run id.
seq	integer	yes	Durable seq of the event (its event-stream id).
sha256	string	yes	SHA-256 of all totalBytes; verify the reassembled text before parsing it.
text	string	yes	Bytes [offset, nextOffset) decoded as UTF-8; a boundary never splits a character.
totalBytes	integer	yes	UTF-8 byte length of the event's complete stream serialization: the redacted JSON line the run event stream delivers.
type	string	yes	Type of an append-only run event, covering run lifecycle, contract/context creation, budget, routing fallbacks, harness activity, interactions, gates, review, arbitration, work products, and control…

No examples provided.

claudexor_run_interactions ~73

List the complete daemon-persisted questions still awaiting answers for a run, including bodies claudexor_run_wait omitted; the text content carries the same JSON.

Input Output Examples

Name	Type	Req	Description
runId	string	yes	Daemon run id (from a run tool's runId trailer or claudexor_runs).

No output schema declared.

No examples provided.

claudexor_run_message ~349

Send a live message to an existing run's active attempt. Requires an explicit stable caller idempotencyKey (the message id); reuse the same key and body to read its recorded receipt. accepted proves native acceptance only (a later live_input_delivered harness event with the same message_id, seen through claudexor_run_wait, reports consumption), delivered proves observed consumption, neither proves obedience. delivery_unknown may have landed: never resend it under a new key. Refusals remain typed outcomes: a pending question answers not_active/interaction_pending (the message does not answer it; use claudexor_answer_interaction), a still-queued job not_active/no_live_session. not_active sent nothing, so a later attempt needs a new key. No automatic retry, queue, or relaunch; host cancellation stops only this request.

Input Output Examples

Name	Type	Req	Description
expectedAttemptId	string	–	Target attempt id. Required when the run has several live attempts (race n>1): without it the daemon answers rejected/multi_attempt. A mismatch with the live attempt answers not_active/attempt_mismat…
idempotencyKey	string	yes	The submission's identity: the Idempotency-Key verbatim. Replaying it returns this receipt again; a new message needs a new key.
runId	string	yes	Run the message was addressed to.
text	string	yes	The message to place into the run's active attempt. 1..65536 UTF-16 code units (the JS string length, not bytes or visible characters); an over-long body is a 400, never truncated. Secret-like values…

Name	Type	Req	Description
accepted	boolean	yes	True only for outcome delivered or accepted; false for every other outcome.
attemptId	string	–	Live attempt the message targeted, when one was selected.
harnessId	string	–	Harness of the targeted attempt.
liveInput	string	–	The targeted harness's declared live-input channel (its capability profile), when an attempt was selected.
message	string	–	Human-readable detail.
messageId	string	yes	The submission's identity: the Idempotency-Key verbatim. Replaying it returns this receipt again; a new message needs a new key.
nativeTurnId	string	–	Native turn id the harness reported on acceptance (Codex turn/steer result).
outcome	string	yes	delivered = native consumption observed; accepted = native acceptance boundary observed (consumption unproved); rejected = explicit refusal of this submission; not_active = no eligible target before…
reason	string	–	Why the outcome is not delivered: no_active_turn (the harness has no active native turn), run_terminal, attempt_mismatch (expectedAttemptId is not the live attempt), no_live_session (no registered li…
runId	string	yes	Run the message was addressed to.

No examples provided.

claudexor_run_result ~63

Read a durable run's terminal result and apply eligibility; non-terminal runs report their current state without pretending to be complete.

Input Output Examples

Name	Type	Req	Description
runId	string	yes	Daemon run id (from a run tool's runId trailer or claudexor_runs).

Name	Type	Req	Description
applyEligibility	–	–	Derived apply-gate verdict; null for read-only runs or when no patch exists.
attemptExecution	array	–	–
budget	–	–	Run budget snapshot (cash + subscription valuation, QA-023c); null when unavailable.
continueFrom	–	–	Predecessor run this run continues; null for an ordinary run, omitted when the detail did not project it.
continuity	array	–	Per-try continuation receipts (carrier, cause, accounts, model) from the run's event log; omitted when not projected.
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan runs.
decisionStatus	string\|null	–	The arbitration decision status, when the run reached arbitration.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	Claudexor Delegate parent id; null for ordinary runs and native subagents.
delegation	–	–	Typed Delegate receipt; null when unavailable or legacy.
detailProblem	–	–	Typed post-terminal detail-read problem (the durable run handle survives); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); null while non-terminal.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges), when terminal.
parentRunId	string\|null	–	Server-owned ordinary parent/follow-up run id, when present.
pendingInteractions	–	–	Count of questions still awaiting answers, when known.
planReadiness	–	–	Derived plan readiness (plan runs only, D17); null otherwise.
resumable	–	–	Terminal facts of stopped unfinished work that claudexor_run_continue can continue (cause, resetsAt, carriers, session, workspace); null when the work finished or the run is not terminal, omitted whe…
retainedEnvelope	–	–	Isolated envelope kept for a successor (disk use visible); null when none, omitted when not projected.
runDir	string\|null	–	Artifact directory; null when absent.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	–	The daemon run id, when known.
secretLike	–	–	INV-062 disclosure: secret-like strings kept in the work but hidden in saved copies (paths and counts only); null when nothing was hidden.
status	string\|null	–	The run's terminal or in-flight lifecycle state, when known.
summary	string	yes	Human-readable outcome/status text (same as the tool's text content).

No examples provided.

claudexor_run_status ~57

Read the current daemon-acknowledged state of a durable Claudexor run.

Input Output Examples

Name	Type	Req	Description
runId	string	yes	Daemon run id (from a run tool's runId trailer or claudexor_runs).

Name	Type	Req	Description
applyEligibility	–	–	Derived apply-gate verdict; null for read-only runs or when no patch exists.
attemptExecution	array	–	–
budget	–	–	Run budget snapshot (cash + subscription valuation, QA-023c); null when unavailable.
continueFrom	–	–	Predecessor run this run continues; null for an ordinary run, omitted when the detail did not project it.
continuity	array	–	Per-try continuation receipts (carrier, cause, accounts, model) from the run's event log; omitted when not projected.
council	–	–	Council membership + merge disclosure (QA-023b); null for solo/non-plan runs.
decisionStatus	string\|null	–	The arbitration decision status, when the run reached arbitration.
deferredMessage	–	–	Compact addressed draft state, with exact predecessor result and successor handles. A terminal predecessor does not imply the draft has resolved.
delegatedFromRunId	string\|null	–	Claudexor Delegate parent id; null for ordinary runs and native subagents.
delegation	–	–	Typed Delegate receipt; null when unavailable or legacy.
detailProblem	–	–	Typed post-terminal detail-read problem (the durable run handle survives); null when the detail read succeeded.
failure	–	–	Typed terminal failure detail and recovery actions; null for non-failed runs.
outcomeBanner	string\|null	–	Server-owned outcome headline (D18); null while non-terminal.
outcomeFacts	–	–	The D8 terminal outcome axes (checks/review/reason/noChanges), when terminal.
parentRunId	string\|null	–	Server-owned ordinary parent/follow-up run id, when present.
pendingInteractions	–	–	Count of questions still awaiting answers, when known.
planReadiness	–	–	Derived plan readiness (plan runs only, D17); null otherwise.
resumable	–	–	Terminal facts of stopped unfinished work that claudexor_run_continue can continue (cause, resetsAt, carriers, session, workspace); null when the work finished or the run is not terminal, omitted whe…
retainedEnvelope	–	–	Isolated envelope kept for a successor (disk use visible); null when none, omitted when not projected.
runDir	string\|null	–	Artifact directory; null when absent.
runFacts	–	–	Validated canonical terminal RunFacts receipt; null while active, unavailable, or for legacy runs.
runId	string\|null	–	The daemon run id, when known.
secretLike	–	–	INV-062 disclosure: secret-like strings kept in the work but hidden in saved copies (paths and counts only); null when nothing was hidden.
status	string\|null	–	The run's terminal or in-flight lifecycle state, when known.
summary	string	yes	Human-readable outcome/status text (same as the tool's text content).

No examples provided.

claudexor_run_time_limit ~98

Read, set, extend, or disable one live run's time limit. Mutations require the current expectedRevision from a read; a conflicting revision must be read again before retrying.

Input Output Examples

Name	Type	Req	Description
action	string	yes	–
expectedRevision	integer	–	–
mode	string	–	For set only; omission keeps the timer's current mode.
runId	string	yes	–
seconds	integer	–	–

No output schema declared.

No examples provided.

claudexor_run_wait ~548

Observe an existing durable run (or a queued job id) through bounded event replay. Defaults: afterSeq=0, waitMs=25000, maxEvents=50, wake=batch. Events already available at the snapshot fence are returned first, even while a question is pending. wake=batch then keeps collecting until the page fills, it includes a question event (also one written before the wait), the run ends, or waitMs elapses; wake=event returns as soon as any event is available. waitMs=0 returns only the already available page. A pending question whose event is not in the page never ends a wait by itself; it is always reported in pendingInteractions. Resume with nextSeq; snapshotSeq is a fence, and terminal is not the end of events, until a page with no events reports eventOmission stream_ended (the stream cannot serve the rest, such as a last line left unterminated by a crash): stop there. Pages are capped at 64 KiB of events, pending bodies at 32 KiB, results at 128 KiB (UTF-8 JSON); an event too large for any page is named in oversizedEvent: read it with claudexor_run_event_read, then resume with afterSeq set to its seq. Daemon discovery and each snapshot read have separate 5s budgets, so a call can take waitMs plus about 15s; size waitMs to your client's tool timeout. Timeout, disconnect and host abort close only this read, never the worker. Use claudexor_run_result for output and claudexor_run_interactions for omitted question bodies.

Input Output Examples

Name	Type	Req	Description
afterSeq	integer	–	Last durable event seq already consumed (the previous nextSeq), never snapshotSeq.
maxEvents	integer	–	–
runId	string	yes	Run id, or the job id of an accepted launch that may still be queued.
waitMs	integer	–	Bound of the event stage. Zero never waits for new events: it returns the bounded page already available at the snapshot fence, read within a separate 5s budget. Daemon discovery and each snapshot re…
wake	string	–	batch (default): collect events until the page fills, it includes a question event (interaction.requested, also one written before the wait), the run ends, or waitMs elapses. event: return as soon as…

Name	Type	Req	Description
deferredMessage	–	–	Snapshot only: terminal wakeReason refers to the predecessor, never proves draft resolution.
eventOmission	–	yes	Why the page stopped early: event_limit or byte_limit (more events may follow; resume with nextSeq), oversized_frame (the next frame exceeds the 131072-byte stream frame cap), invalid_event (the next…
events	array	yes	Ordered new events, at most 65536 UTF-8 JSON bytes including envelopes. Legacy events retain their body and use the stream id as seq.
nextSeq	integer	yes	Last returned event seq, or afterSeq when none was returned. It never advances over an unreturned event.
omittedInteractions	integer	yes	–
oversizedEvent	–	yes	The next unread event after nextSeq cannot fit one page. Read its exact bytes in parts with claudexor_run_event_read, then resume with afterSeq set to its seq. Null otherwise.
pendingInteractionCount	integer	yes	–
pendingInteractions	array	yes	Current snapshot's pending questions, bounded to 32768 UTF-8 JSON bytes; claudexor_run_interactions returns complete bodies. They never hold back events.
runId	string	yes	Canonical run id; a queued job id converges to its run id once the run binds.
snapshotPhase	string	yes	–
snapshotProblem	–	yes	A failed final read leaves the explicitly labelled initial snapshot, never invented current state.
snapshotSeq	integer	yes	Last event seq at the latest snapshot: a fence, not a cursor. nextSeq below it means events are already written; eventOmission stream_ended reports when the stream cannot serve them.
status	string	yes	Control-plane run state = the run LIFECYCLE (D8): queued/running while live, then exactly one of succeeded / failed / cancelled / interrupted. Outcome quality (checks/review/delivery) is orthogonal —…
wakeReason	string	yes	Why this observation returned, by precedence: aborted (this read was cancelled; the worker continues), terminal (the run is terminal; events may still follow nextSeq unless eventOmission is stream_en…

No examples provided.

claudexor_runs ~31

List recent daemon-tracked Claudexor runs (recovery: find a lost runId).

Input Output Examples

Input schema present but exposes no named parameters.

No output schema declared.

No examples provided.

claudexor_runs_wait ~215

Observe lifecycle and pending attention from 1..1000 existing durable job/run handles in one read. targets contains {runId, seen?}; waitMs defaults to 25000 (0..50000). Echo each summary token as that target's seen to suppress unchanged terminal/question attention; tokens are not event cursors or answers. Returns ascending attention indexes and compact summaries in input order, including duplicates and explicit unknowns, within 128 KiB of logical JSON. waitMs=0 reads once; timeout never retries. Discovery adds up to 5s and HTTP adds up to waitMs+5s (total waitMs+10s). Host abort closes only observation; discovery's read-only handshake may settle in the background. Requires an existing supporting daemon; never starts one or a worker. Terminal lifecycle is not output readiness: use claudexor_run_result for output and claudexor_run_interactions for question bodies.

Input Output Examples

Name	Type	Req	Description
targets	array	yes	–
waitMs	integer	–	–

Name	Type	Req	Description
attention	array	yes	–
summaries	array	yes	–
wakeReason	string	yes	–

No examples provided.

claudexor_status ~42

Return doctor-backed Claudexor runtime status: per-harness verdicts with enabled/disabled intents, doctor reasons and checks, and the configured model.

Input Output Examples

Input schema present but exposes no named parameters.

No output schema declared.

No examples provided.

claudexor_store_identity ~73

Read the running daemon's durable store identity (storeId) through the opt-in handshake. Save storeId durably with each import key BEFORE claudexor_upload_import, and replay that saved value unchanged; never replace it with a later answer. Resource ids and import keys belong to that store only.

Input Output Examples

Input schema present but exposes no named parameters.

Name	Type	Req	Description
compatible	boolean	yes	–
engine	object	yes	Build identity of the serving engine.
operationsPath	string	yes	–
protocolMajor	number	yes	–
servingMode	string	–	Product admission: 'recovery_only' keeps normal product routes closed while recovery control stays reachable; absent means 'normal'.
storeId	string	–	Immutable UUID of the durable engine store, stable across restarts; a run 404 is meaningful only with the same storeId. Returned only to a request sending `X-Claudexor-Store-Identity: 1` (older stric…

No examples provided.

claudexor_task_group_runs ~292

Read exactly one retained task group page by caller UUID, optionally checking expectedStoreId. limit is 1..200 (default 50); state filters lifecycle; cursor (at most 2048 characters) is bound to group/filter/store/current served generations. The logical result is at most 131072 UTF-8 bytes and may return fewer members; resume only with nextCursor. Retained-only/current-served qualifiers are not a permanent roster or proof of submission. Job handles remain canonical while runId is null until binding. Explicit unavailable observations are unknown. Requires an existing supporting daemon; discovery and HTTP reads each allow up to 5s. Host cancellation closes only the read. Never launches, retries, resumes, cancels or answers work. Lifecycle is not output readiness; use addressed result/interactions tools for bodies.

Input Output Examples

Name	Type	Req	Description
cursor	string	–	–
expectedStoreId	string	–	–
limit	integer	–	–
state	string	–	Control-plane run state = the run LIFECYCLE (D8): queued/running while live, then exactly one of succeeded / failed / cancelled / interrupted. Outcome quality (checks/review/delivery) is orthogonal —…
taskGroupId	string	yes	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.

Name	Type	Req	Description
hasMore	boolean	yes	–
members	array	yes	–
nextCursor	–	yes	–
retention	string	yes	–
storeId	string	yes	–
taskGroupId	string	yes	Caller-generated UUID saved before launching the first member. Each job still needs its own idempotency key; the group grants no execution authority.
viewToken	string	yes	–
visibility	string	yes	–

No examples provided.

Install

How do I install the Claudexor MCP server?

Claudexor runs locally as an npm package, launched with npx -y claudexor. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

npm · claudexor

Client

Showing Claude. Snippets for every client (markdown)

# add to Claude Code

claude mcp add razzant-claudexor -- npx -y claudexor

Common questions

What is the Claudexor MCP server?

Claudexor is an MCP server listed in the public MCP registry as io.github.razzant/claudexor. Route coding work across Claude Code, Codex, Cursor, and OpenCode with shared context. This page covers its npm package (claudexor).

Is the Claudexor MCP server safe to use?

Claudexor scores 89 out of 100 on VerifyMCP. We recorded 32 known advisories against it as of 4 November 2026. It declares no install or post-install scripts. Its build provenance is signed and verified. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

What tools does the Claudexor MCP server expose?

Claudexor exposes 41 tools: claudexor_ask, claudexor_run, claudexor_best_of, claudexor_plan, claudexor_create, and 36 more. Their descriptions and schemas cost roughly 8,811 tokens of context every time the server is loaded.

Is the Claudexor MCP server still maintained?

Claudexor is still listed as active in the MCP registry. We last reached this channel on 4 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

What licence is the Claudexor MCP server under?

Claudexor declares the MIT licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

On this page

Handy links

Scoring key

80+ high trust
60–79 medium
below 60 low

How we score

Badge

Put this score in your README. The image is served from a stable URL, updates when the score does, and links back to the evidence on this page. It shows nothing that is not already published here.

Score Verified Plain

The two-cell badge: our name, and this channel's current score out of 100.

Markdown

 [![VerifyMCP trust score for Claudexor](https://verifymcp.io/badge/razzant-claudexor/claudexor.svg)](https://verifymcp.io/servers/razzant-claudexor/claudexor)

HTML

 <a href="https://verifymcp.io/servers/razzant-claudexor/claudexor"><img src="https://verifymcp.io/badge/razzant-claudexor/claudexor.svg" alt="VerifyMCP trust score for Claudexor" height="20"></a>

A single cell stating that the server is listed here. Carries no number.

Markdown

 [![VerifyMCP trust score for Claudexor](https://verifymcp.io/badge/razzant-claudexor/claudexor.svg?variant=verified)](https://verifymcp.io/servers/razzant-claudexor/claudexor)

HTML

 <a href="https://verifymcp.io/servers/razzant-claudexor/claudexor"><img src="https://verifymcp.io/badge/razzant-claudexor/claudexor.svg?variant=verified" alt="VerifyMCP trust score for Claudexor" height="20"></a>

The score alone, tier-coloured, for a badge row that is already crowded.

Markdown

 [![VerifyMCP trust score for Claudexor](https://verifymcp.io/badge/razzant-claudexor/claudexor.svg?variant=plain)](https://verifymcp.io/servers/razzant-claudexor/claudexor)

HTML

 <a href="https://verifymcp.io/servers/razzant-claudexor/claudexor"><img src="https://verifymcp.io/badge/razzant-claudexor/claudexor.svg?variant=plain" alt="VerifyMCP trust score for Claudexor" height="20"></a>

Score history · last 30 days

12 Oct 4 Nov

Trust score by day, last 30 days
Day	Score
12 Oct	47
13 Oct	47
14 Oct	47
15 Oct	90
16 Oct	90
17 Oct	90
18 Oct	90
19 Oct	90
20 Oct	90
21 Oct	90
22 Oct	90
23 Oct	90
24 Oct	90
25 Oct	90
26 Oct	90
27 Oct	90
28 Oct	89
29 Oct	89
30 Oct	89
31 Oct	89
1 Nov	89
2 Nov	89
3 Nov	89
4 Nov	89