# subtask (pypi · subtask-mcp)

Background sub-agents for Claude Code, Codex, Kimi and OpenCode, over MCP, with a spend cap.

- Trust score: 81/100 (high trust)
- Change this week: +18
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-04

## Components

- pypi · `subtask-mcp`: 81/100 (this document), [markdown](https://verifymcp.io/servers/tedin7-subtask/subtask-mcp.md), [page](https://verifymcp.io/servers/tedin7-subtask/subtask-mcp)

## Channel facts

- Registry: `pypi`
- Package: `subtask-mcp`
- Version: `0.5.9`
- Transport: `stdio`

## Trust breakdown

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. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-11-04.

- **Supply Chain Security**: 100/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - Runs setuptools.build_meta at install time, a recognised build step with no custom scripting around it.
  - 2 of 37 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 100/100
  - Source repository is publicly reachable at the declared URL.
  - Cryptographically verified build provenance (signed, bound to tedin7/subtask).
  - Clear OSI-approved license (MIT).
  - Actively maintained (last published 0 days ago).
  - Publishes a security disclosure policy (SECURITY.md).
- **Schema Quality & AI Usability**: 51/100
  - AI-judged instruction clarity (fair).
  - Context-footprint check failed: tool/resource definitions use about 2378 tokens (~132/item across 18 items; 18 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 57/100
  - Stability check failed: the tool surface changed between 0.2.11 and 0.5.9: 1 tool removals, 0 breaking changes, 9 additions.
- **Tool Coverage**: 63/100
  - 94% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 0% of tool parameters carry a description.
- **Tool Safety**: 75/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - 0 of 1 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation; "cleanup" implies "remove" and declares no destructiveHint at all, which the MCP spec reads as destructive by default.
  - An AI judge read all 19 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a current MCP spec version (2026-07-28).

## Install

### How do I install the subtask MCP server?

subtask runs locally as a PyPI package, launched with uvx subtask-mcp. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add tedin7-subtask -- uvx subtask-mcp
```

### Cursor

```json
{
  "mcpServers": {
    "tedin7-subtask": {
      "command": "uvx",
      "args": [
        "subtask-mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "tedin7-subtask": {
      "command": "uvx",
      "args": [
        "subtask-mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add tedin7-subtask -- uvx subtask-mcp
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "tedin7-subtask": {
      "type": "local",
      "command": [
        "uvx",
        "subtask-mcp"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add tedin7-subtask --command uvx --arg subtask-mcp
```

### Hermes

```yaml
mcp_servers:
  tedin7-subtask:
    command: "uvx"
    args: ["subtask-mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "tedin7-subtask": {
      "Transport": "stdio",
      "Command": "uvx",
      "Arguments": [
        "subtask-mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add tedin7-subtask -t stdio -c uvx -a subtask-mcp
```

### Other

```json
{
  "mcpServers": {
    "tedin7-subtask": {
      "command": "uvx",
      "args": [
        "subtask-mcp"
      ]
    }
  }
}
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-11-04 (score 81, +16)

- [security improvement] Malware scan: unverified → pass

### 2026-11-03 (score 65, 0)

- [security improvement] Malware scan: unverified → pass
- [functional] Package version: 0.5.5 → 0.5.9
- [functional] Package version: 0.5.5 → 0.5.8
- [functional] Package version: 0.5.5 → 0.5.6

### 2026-11-02 (score 65, −14)

- [security regression] Malware scan: pass → unverified

### 2026-11-01 (score 79, +15)

- [security regression] Stability: fail → unverified
- [security regression] Tool safety: pass → unverified
- [security improvement] Malware scan: unverified → pass
- [functional regression] Schema quality: 107 → 132
- [functional regression] Capabilities: pass → unverified
- [functional regression] Tool coverage: 93 → unverified
- [functional] Package version: 0.5.0 → 0.5.5
- [functional] Package version: 0.5.0 → 0.5.4
- [functional] Package version: 0.5.0 → 0.5.3

### 2026-10-31 (score 64, −14)

- [security regression] Malware scan: pass → unverified
- [functional improvement] Security disclosure: unverified → pass

### 2026-10-30 (score 78, −1)

- [functional regression] Security disclosure: pass → unverified

### 2026-10-29 (score 79, +16)

- [security improvement] Malware scan: unverified → pass

### 2026-10-27 (score 63, +1)

No change was recorded against any check on this day. Stability & Change Management went from 27 to 30.

## MCP tools (18)

### `plan` (~97 tokens)

Preview tasks before workers start. Read deterministic facts and advisory Jev contradictions, then
launch=True with plan_id to start the unchanged plan once. Changed inputs require a new preview.
This is a coordinator decision within the authorized task, not a new operator permission.
Legacy start/start_many remain available. Read dependencies must be explicit in depends_on.

Input parameters:

- `launch` (boolean)
- `plan_id` (string)
- `tasks` (array)

### `start` (~436 tokens)

Start one job; returns its id. kind: read|code|review|hard. Writes: repo= or isolation='worktree' (own branch).
model: engine-specific model id (opencode, codex, kimi, API engines); omitted = the engine's default.
verify: shell check run by subtask after the job; max_retries (0-5) resumes with the failure.
invariants: default ['no_test_deletion','no_suppression'] fail verify on deleted/skipped tests or new noqa; [] = off.
allowed_paths: files outside are reverted. symbol_scope: ['f.py::func'] only those defs may change; integrate merges by symbol.
oracle_cmd: after verify passes, run on base and result; reports output_match and speedup_pct.
adversarial: a Breaker job writes the tests first. context_refs: file:P[:A-B] | git:diff[:REF] | subtask:ID:summary|diff
instead of pasting text; auto_repomap adds a symbol map. policy: worktree|sandbox|isolated|untrusted. tier: auto|fast|balanced|deep|free_only.

Input parameters:

- `adversarial` (boolean)
- `allowed_paths` (array)
- `auto_repomap` (boolean)
- `base` (string)
- `context_refs` (array)
- `contract` (object)
- `cwd` (string)
- `effort` (string)
- `engine` (string)
- `files` (array)
- `invariants` (array)
- `isolation` (string)
- `kind` (string)
- `max_retries` (integer)
- `max_turns` (integer)
- `max_usd` (number)
- `model` (string)
- `oracle_cmd` (string)
- `policy` (string)
- `readonly` (boolean)
- `repo` (string)
- `symbol_scope` (array)
- `task` (string, required)
- `tier` (string)
- `verify` (string)

### `start_many` (~140 tokens)

Start several jobs in one call. Items take start()'s arguments. With id + depends_on they form a group (grp-…)
run as a DAG; subtask:<id>:summary in task/context_refs is replaced by the real job id.
With Jev auto, the response includes advisory dependency checks; read them without treating them as permission.
foreach=<parent id>: one copy per item of the JSON list the parent writes to .subtask/fanout.json ({{item}} in
task), max 8; dependents wait for every copy. after='any': start when one parent succeeded.

Input parameters:

- `tasks` (array, required)

### `wait` (~198 tokens)

Block until jobs or groups end and return their summaries. mode: all | any (first to end) |
first_success (first done and verified; the rest are stopped). Returns early on a stalled job.
since: cursor for clients without progress; with it, timeout is capped at 60 s. For Codex:
wait(ids, mode='any', timeout=60, since=''), drop ended ids and pass back cursor=. An idle Codex
0.164.0 is not woken by ordinary local MCP progress or log notifications. Includes the automatic
Jev delivery/evidence/risk advice, raw provider choice, wrapper assessment and next action:
read it before a manual review and do not resubmit the same work.

Input parameters:

- `any` (boolean)
- `ids` (array, required)
- `mode` (string)
- `since` (string)
- `stall_timeout` (number)
- `timeout` (integer)

### `status` (~19 tokens)

One line per job.

Input parameters:

- `id` (string)

### `result` (~43 tokens)

Job summary; full=True for all; evidence=True for the JSON envelope.

Input parameters:

- `evidence` (boolean)
- `full` (boolean)
- `id` (string, required)

### `reply` (~38 tokens)

Jev auto adds advisory feedback checks; read the output. Continue a finished job.

Input parameters:

- `id` (string, required)
- `message` (string, required)

### `stop` (~18 tokens)

Stop a job.

Input parameters:

- `id` (string, required)

### `steer` (~52 tokens)

Correct a job: hint (instruction sent when the attempt ends) | abort | retry (fresh job with the instruction).

Input parameters:

- `action` (string, required)
- `id` (string, required)
- `instruction` (string)

### `jev` (~138 tokens)

Show or change the shared Jev policy: status | auto | manual | off. Only the operator changes
the policy: do not switch it to obtain or avoid a check. Auto checks every new final delivery once
within the existing budget; read that outcome in wait()/result before any manual review. All advice
is consultative. Credentials are never accepted here: add or replace a key in a terminal with
\`subtask jev setup` (hidden input), never in chat, argv or a tool argument. Changes apply to new
submissions in every configured CLI; no server reload is needed.

Input parameters:

- `action` (string)

### `review` (~229 tokens)

Advisory Jev review of a terminal delivery against the parent's criteria (1-12). Read the
automatic outcome in wait()/result first; do not resubmit the same delivery. Default prepares a
local evidence packet without network; packets=True prepares separate claim packets and manifests
offline (never submits). submit=True sends one request (unknown sends are never
retried). Supply original_request from the coordinator to judge full scope; without it
scope_assessment is unproven, which is neither a pass nor a contradiction. supported is advice,
not acceptance. provider_choice is the raw answer; assessment includes local vetoes (missing,
failed or mismatched evidence), which are not findings by the model. Jev never accepts, merges,
deploys or retries the worker; local checks remain necessary. Key/budget errors are final for the
call: keys only from `subtask jev setup`, never a tool argument.

Input parameters:

- `criteria` (array, required)
- `id` (string, required)
- `original_request` (string)
- `packets` (boolean)
- `submit` (boolean)

### `inspect` (~52 tokens)

Raw log lines of a job (last `tail`, filtered by `grep`). Only when the summary is not enough.

Input parameters:

- `grep` (string)
- `id` (string, required)
- `tail` (integer)

### `integrate` (~79 tokens)

Bring worktree jobs back. diff first to review; apply = uncommitted, all or nothing, Python files of
symbol_scope jobs merged by symbol; discard removes worktree and branch. auto_resolve: a reconcile job on conflict.

Input parameters:

- `action` (string)
- `auto_resolve` (boolean)
- `ids` (array, required)

### `bisect` (~85 tokens)

First commit in good_ref..bad_ref that fails test_cmd, testing up to parallelism commits at once. auto_fix: start a fix job.

Input parameters:

- `auto_fix` (boolean)
- `bad_ref` (string, required)
- `good_ref` (string, required)
- `parallelism` (integer)
- `repo` (string)
- `test_cmd` (string, required)

### `consensus` (~83 tokens)

2-4 engines solve prompt blind in worktrees, then one job merges the best (verify, invariants). Returns a group id for wait().

Input parameters:

- `engines` (array, required)
- `isolation` (string)
- `max_usd` (number)
- `prompt` (string, required)
- `repo` (string)
- `verify_cmd` (string)

### `map_reduce` (~123 tokens)

The same edit over many files: one job per batch of files matching glob ({{files}} in prompt_template),
at most `concurrency` at once, each in a worktree with verify_cmd. Returns a group id for wait().

Input parameters:

- `batch_size` (integer)
- `concurrency` (integer)
- `cwd` (string)
- `engine` (string)
- `glob_pattern` (string, required)
- `isolation` (string)
- `prompt_template` (string, required)
- `repo` (string)
- `verify_cmd` (string)

### `cleanup` (~24 tokens)

Remove a job's worktree, keep branch.

Input parameters:

- `id` (string, required)

### `engines` (~13 tokens)

Free quota and API spend.

## Diagnostics

Captured diagnostic sections: Provenance, Install scripts, Dependencies. The full working is on the page: https://verifymcp.io/servers/tedin7-subtask/subtask-mcp#diagnostics

## Score history

- 2026-11-04: 81
- 2026-11-03: 65
- 2026-11-02: 65
- 2026-11-01: 79
- 2026-10-31: 64
- 2026-10-30: 78
- 2026-10-29: 79
- 2026-10-28: 63
- 2026-10-27: 63
- 2026-10-26: 62
- 2026-10-25: 62
- 2026-10-24: 61
- 2026-10-23: 62
- 2026-10-22: 76
- 2026-10-21: 61
- 2026-10-20: 76
- 2026-10-19: 75
- 2026-10-18: 59
- 2026-10-17: 73
- 2026-10-16: 57
- 2026-10-15: 72
- 2026-10-14: 50

## Common questions

### What is the subtask MCP server?

subtask is an MCP server listed in the public MCP registry as io.github.tedin7/subtask. Background sub-agents for Claude Code, Codex, Kimi and OpenCode, over MCP, with a spend cap. This page covers its PyPI package (subtask-mcp).

### Is the subtask MCP server safe to use?

subtask scores 81 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 4 November 2026. 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 subtask MCP server expose?

subtask exposes 18 tools: plan, start, start_many, wait, status, and 13 more. Their descriptions and schemas cost roughly 1,867 tokens of context every time the server is loaded.

### Is the subtask MCP server still maintained?

subtask 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 subtask MCP server under?

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

## Links

- PyPI project: https://pypi.org/project/subtask-mcp/
- Socket report: https://socket.dev/pypi/package/subtask-mcp
- Repository: https://github.com/tedin7/subtask
- Changelog RSS feed: https://verifymcp.io/servers/tedin7-subtask/subtask-mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/tedin7-subtask/subtask-mcp.json
- HTML version of this page: https://verifymcp.io/servers/tedin7-subtask/subtask-mcp