# DidWork (npm · @didwork/mcp)

Verify that an agent's claimed outcome actually happened, using independently gathered evidence

- Trust score: 73/100 (medium)
- Change this week: +2
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-09-23

## Components

- npm · `@didwork/mcp`: 73/100 (this document), [markdown](https://verifymcp.io/servers/sh-didwork-didwork/didwork-mcp.md), [page](https://verifymcp.io/servers/sh-didwork-didwork/didwork-mcp)

## Channel facts

- Registry: `npm`
- Package: `@didwork/mcp`
- Version: `0.1.1`
- 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 2026-09-23.

- **Supply Chain Security**: 98/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - No install/post-install scripts declared.
  - 32 of 97 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 19/100
  - Repository check failed: no source repository is declared.
  - Provenance check failed: no build-provenance attestation is published.
  - Clear OSI-approved license (Apache-2.0).
  - Actively maintained (last published 0 days ago).
  - Security-disclosure policy not yet verified: we couldn't inspect the source repository.
- **Schema Quality & AI Usability**: 64/100
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 1835 tokens (~183/item across 10 items; 10 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 80/100
  - Stability observed for 24 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 87/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 60% of tool parameters carry a description.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 10 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 10 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

## Install

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

DidWork runs locally as an npm package, launched with npx -y @didwork/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 sh-didwork-didwork -- npx -y @didwork/mcp
```

### Cursor

```json
{
  "mcpServers": {
    "sh-didwork-didwork": {
      "command": "npx",
      "args": [
        "-y",
        "@didwork/mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "sh-didwork-didwork": {
      "command": "npx",
      "args": [
        "-y",
        "@didwork/mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add sh-didwork-didwork -- npx -y @didwork/mcp
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "sh-didwork-didwork": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "@didwork/mcp"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add sh-didwork-didwork --command npx --arg -y --arg @didwork/mcp
```

### Hermes

```yaml
mcp_servers:
  sh-didwork-didwork:
    command: "npx"
    args: ["-y", "@didwork/mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "sh-didwork-didwork": {
      "Transport": "stdio",
      "Command": "npx",
      "Arguments": [
        "-y",
        "@didwork/mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add sh-didwork-didwork -t stdio -c npx -a -y @didwork/mcp
```

### Other

```json
{
  "mcpServers": {
    "sh-didwork-didwork": {
      "command": "npx",
      "args": [
        "-y",
        "@didwork/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.

### 2026-09-23 (score 73, −1)

- [functional regression] Schema quality: 134 → 183
- [functional] Package version: 0.1.0 → 0.1.1

### 2026-09-22 (score 74, +1)

No change was recorded against any check on this day. Stability & Change Management went from 73 to 77. That category is still filling its 30-day observation window: 22 days of observed history at the previous scan, 23 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-21 (score 73, 0)

- [functional regression] Tool coverage: 67% → 60%
- [functional regression] Schema quality: 1120 → 1344
- [functional regression] Security disclosure: fail → unverified
- [functional] Package version: 0.0.10 → 0.1.0

### 2026-09-20 (score 73, +1)

No change was recorded against any check on this day. Stability & Change Management went from 67 to 70. That category is still filling its 30-day observation window: 20 days of observed history at the previous scan, 21 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-17 (score 72, +1)

No change was recorded against any check on this day. Stability & Change Management went from 57 to 60. That category is still filling its 30-day observation window: 17 days of observed history at the previous scan, 18 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-15 (score 71, +1)

No change was recorded against any check on this day. Stability & Change Management went from 50 to 53. That category is still filling its 30-day observation window: 15 days of observed history at the previous scan, 16 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-13 (score 70, +1)

No change was recorded against any check on this day. Stability & Change Management went from 43 to 47. That category is still filling its 30-day observation window: 13 days of observed history at the previous scan, 14 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-11 (score 69, +1)

No change was recorded against any check on this day. Stability & Change Management went from 37 to 40. That category is still filling its 30-day observation window: 11 days of observed history at the previous scan, 12 at this one. The score rises as the window fills, whether or not the server changes.

## MCP tools (10)

### `did_verify` (~727 tokens)

Independently verify that an intended outcome is actually true. Sends claim fields (including private repository identifiers, branches and commit SHAs), optional subject/account labels and tool profiles to the configured DidWork API (default https://api.didwork.sh). DidWork reads provider state without modifying it, but authenticated verification creates a DidWork record containing the claim, verdict, timestamps and evidence, which may include provider response data. Verification evidence is retained under the account plan (Free: 7 days; Pro: 365 days). Keyless http.ok does not store a verification record. Host approval rejection means the tool was blocked, not a failed or unknown DidWork verdict; report it separately and do not retry through another route. Send a claim (type + expected); DidWork gathers evidence from the authoritative system and returns a verdict: verified, failed, or unknown. Gate your next action on the verdict — an agent should not grade its own work. Treat unknown as 'do not proceed yet'. Prefer the most specific claim type for the outcome; http.ok is the fallback and is unauthenticated, so it cannot see private repos, dashboards or anything behind a login — against those it reports a 404 as failed, which says nothing about whether the work landed. Claim types, by provider (connect providers at https://didwork.sh/console): stripe.refund, stripe.payment_succeeded, stripe.subscription_active, stripe.subscription_cancelled, stripe.invoice_paid, stripe.checkout_completed, stripe.payout_paid, stripe.payment_method_attached; github.pr_merged, github.workflow_passed, github.issue_closed, github.release_published, github.commit_in_branch, github.file_exists, github.deployment_succeeded, github.pr_review_approved, github.branch_exists; gitlab.mr_merged, gitlab.pipeline_passed, gitlab.issue_closed, gitlab.commit_in_branch, gitlab.file_exists, gitlab.release_published; linear.issue_completed, linear.issue_in_state, linear.issue_assigned; jira.issue_done,…

Input parameters:

- `account` (string): Which connected account to verify against, by the label the user gave it in the console (e.g. 'acme'). Omit to try every account connected for the provider.
- `expected` (object, required): The claim's expected fields. For github.commit_in_branch use { repository: 'owner/repo', branch: 'main', commit: 'full-sha' }; for stripe.refund use { payment: 'pi_123', amount: 4999 }.
- `subject` (string): Optional caller-side reference id
- `tools` (array): The tools that performed the work, as profiles from did_inspect_tool or { name, capabilities } records. Recorded on the receipt as caller-reported context (what the tools could affect); the verdict d…
- `type` (string, required): Claim type, e.g. 'stripe.refund' or 'http.ok'
- `wait` (boolean): false = return a pending verification immediately and poll with did_get

### `did_get` (~28 tokens)

Fetch a verification by id (poll a pending async verification).

Input parameters:

- `verification_id` (string, required)

### `did_verify_contract` (~257 tokens)

Verify a reviewed version 2 outcome contract. Pass the requirements unchanged; never remove or weaken requirements to obtain a pass. The caller must load the contract from the trusted source. DidWork records caller-supplied provenance, independently gathers claim evidence and returns a run. Reuse the same idempotency_key after an interrupted request. A verified run covers only its required checks. Requires an API key. Sends claim fields (including private repository identifiers, branches and commit SHAs), optional subject/account labels and tool profiles to the configured DidWork API (default https://api.didwork.sh). DidWork reads provider state without modifying it, but authenticated verification creates a DidWork record containing the claim, verdict, timestamps and evidence, which may include provider response data. Verification evidence is retained under the account plan (Free: 7 days; Pro: 365 days). Keyless http.ok does not store a verification record. Host approval rejection means the tool was blocked, not a failed or unknown DidWork verdict; report it separately and do not retry through another route. The full contract and caller-supplied provenance are also sent and stored on the outcome run.

Input parameters:

- `contract` (object, required)
- `idempotency_key` (string, required)

### `did_get_run` (~34 tokens)

Read an outcome run by id. Running, failed and unknown are not completed success.

Input parameters:

- `run_id` (string, required)

### `did_list` (~30 tokens)

Recent verifications for this API key — the developer log, evidence included.

Input parameters:

- `limit` (integer)

### `did_watch` (~410 tokens)

Continuously re-verify a claim on an interval (e.g. '10m', '1h'). DidWork notifies the webhook_url on the first run and on verdict transitions. Plan limits apply: a free key may run one watch, hourly at most ('1h'), and a shorter interval or a second watch is refused with 402 watch_cadence_requires_pro or 402 watch_limit_reached. Pro keys run unlimited watches down to '5m'. If a watch is refused, say so — do not retry at the same cadence. Sends claim fields (including private repository identifiers, branches and commit SHAs), optional subject/account labels and tool profiles to the configured DidWork API (default https://api.didwork.sh). DidWork reads provider state without modifying it, but authenticated verification creates a DidWork record containing the claim, verdict, timestamps and evidence, which may include provider response data. Verification evidence is retained under the account plan (Free: 7 days; Pro: 365 days). Keyless http.ok does not store a verification record. Host approval rejection means the tool was blocked, not a failed or unknown DidWork verdict; report it separately and do not retry through another route. Stores the schedule and webhook URL in DidWork; sends notifications to that URL.

Input parameters:

- `account` (string): Which connected account to verify against, by the label the user gave it in the console (e.g. 'acme'). Omit to try every account connected for the provider.
- `every` (string): Interval like '10m', '1h', '1d' (default 1h)
- `expected` (object, required): The claim's expected fields. For github.commit_in_branch use { repository: 'owner/repo', branch: 'main', commit: 'full-sha' }; for stripe.refund use { payment: 'pi_123', amount: 4999 }.
- `type` (string, required)
- `webhook_url` (string)

### `did_watches` (~18 tokens)

List active watches with their last verdicts.

### `did_unwatch` (~23 tokens)

Stop a watch by id.

Input parameters:

- `watch_id` (string, required)

### `did_usage` (~15 tokens)

Verifications performed, by month.

### `did_inspect_tool` (~293 tokens)

Capability Trust: before calling an agent tool with real-world consequences (money, email, deploys, deletes, permissions), ask what it can actually affect. DidWork statically analyses the tool implementations in a project — MCP server.tool / registerTool handlers and did.tool() declarations, TypeScript/JavaScript only — and returns a capability profile per tool: evidenced capabilities (e.g. financial.refund, communication.external, data.delete) each backed by file:line evidence, a risk level (low/moderate/high/critical/unknown), a separate confidence level, and a status: verified (profile established), failed (DECLARATION_MISMATCH: the implementation exceeds what did.tool() declares), or unknown (insufficient evidence). Treat unknown as 'not shown to be safe', never as low risk; it usually means the tool calls something injected at runtime or loads code dynamically. Omit `tool` for the project summary and one row per tool; pass `tool` for the full profile with evidence and gaps. Runs locally, reads source only, no API key needed, nothing is uploaded.

Input parameters:

- `max_depth` (integer): Project-internal calls to follow from each handler (default 8)
- `root` (string): Project directory to inspect (default: the directory the host launched the server in)
- `tool` (string): Tool name for the full profile; omit for the summary

## Diagnostics

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

## Score history

- 2026-09-23: 73
- 2026-09-22: 74
- 2026-09-21: 73
- 2026-09-20: 73
- 2026-09-19: 72
- 2026-09-18: 72
- 2026-09-17: 72
- 2026-09-16: 71
- 2026-09-15: 71
- 2026-09-14: 70
- 2026-09-13: 70
- 2026-09-12: 69
- 2026-09-11: 69
- 2026-09-10: 68
- 2026-09-09: 68
- 2026-09-08: 67
- 2026-09-07: 67
- 2026-09-06: 66
- 2026-09-05: 51
- 2026-09-04: 33
- 2026-09-03: 66
- 2026-09-02: 65
- 2026-09-01: 65
- 2026-08-31: 64
- 2026-08-30: 49

## Common questions

### What is the DidWork MCP server?

DidWork is an MCP server listed in the public MCP registry as sh.didwork/didwork. Verify that an agent's claimed outcome actually happened, using independently gathered evidence. This page covers its npm package (@didwork/mcp).

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

DidWork scores 73 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 23 September 2026. It declares no install or post-install scripts. 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 DidWork MCP server expose?

DidWork exposes 10 tools: did_verify, did_get, did_verify_contract, did_get_run, did_list, and 5 more. Their descriptions and schemas cost roughly 1,835 tokens of context every time the server is loaded.

### Is the DidWork MCP server still maintained?

DidWork is still listed as active in the MCP registry. We last reached this channel on 23 September 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 DidWork MCP server under?

DidWork declares the Apache-2.0 licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

## Links

- npm package: https://www.npmjs.com/package/@didwork/mcp
- Socket report: https://socket.dev/npm/package/@didwork/mcp
- Website: https://didwork.sh/
- Changelog RSS feed: https://verifymcp.io/servers/sh-didwork-didwork/didwork-mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/sh-didwork-didwork/didwork-mcp.json
- HTML version of this page: https://verifymcp.io/servers/sh-didwork-didwork/didwork-mcp