# io.github.Garl-Protocol/agent-trust (npm · @garl-protocol/mcp-server)

Tamper-evident action receipts, trust scoring & capability tokens for AI agents. 29 MCP tools.

- Trust score: 62/100 (medium)
- Change this week: +16
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-08-03

## Components

- npm · `@garl-protocol/mcp-server`: 62/100 (this document), [markdown](https://verifymcp.io/servers/garl-protocol-agent-trust/garl-protocol-mcp-server.md), [page](https://verifymcp.io/servers/garl-protocol-agent-trust/garl-protocol-mcp-server)

## Channel facts

- Registry: `npm`
- Package: `@garl-protocol/mcp-server`
- Version: `1.4.3`
- 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 2026-08-03.

- **Supply Chain Security**: 86/100
  - No malware found by supply-chain analysis.
  - Only part of the dependency tree could be resolved (94 of 98), so this covers what we could see, not the whole tree.
  - No install/post-install scripts declared.
  - Only part of the dependency tree could be resolved (94 of 98), so this covers what we could see, not the whole tree.
- **Provenance & Transparency**: 45/100
  - Source repository is publicly reachable at the declared URL.
  - Provenance check failed: no build-provenance attestation is published.
  - Clear OSI-approved license (Apache-2.0).
  - Actively maintained (last published 56 days ago).
  - Disclosure check failed: no security disclosure policy was found in the source repository.
- **Schema Quality & AI Usability**: 43/100
  - AI-judged instruction clarity (poor).
  - Tool/resource definitions use about 2538 tokens (~87/item across 29 items; 29 tools + 0 resources), lean.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 27/100
  - Stability observed for 8 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 91/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 74% of tool parameters carry a description.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

## Install

### Claude

```bash
claude mcp add garl-protocol-agent-trust -- npx -y @garl-protocol/mcp-server
```

### Codex

```bash
codex mcp add garl-protocol-agent-trust -- npx -y @garl-protocol/mcp-server
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "garl-protocol-agent-trust": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "@garl-protocol/mcp-server"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add garl-protocol-agent-trust --command npx --arg -y --arg @garl-protocol/mcp-server
```

### Hermes

```yaml
mcp_servers:
  garl-protocol-agent-trust:
    command: "npx"
    args: ["-y", "@garl-protocol/mcp-server"]
```

### Other

```json
{
  "mcpServers": {
    "garl-protocol-agent-trust": {
      "command": "npx",
      "args": [
        "-y",
        "@garl-protocol/mcp-server"
      ]
    }
  }
}
```

## 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.

### 2026-08-03 (score 62, +42)

- [security regression] Provenance: unverified → fail
- [security improvement] Known CVEs: unverified → partial
- [security improvement] Install scripts: unverified → pass
- [functional regression] Tool coverage: 100% → 74%
- [functional regression] Schema quality: 66 → 87
- [functional regression] Security disclosure: unverified → fail
- [functional improvement] Schema quality: unverified → poor
- [functional improvement] Tool coverage: unverified → 100
- [functional improvement] License: unverified → pass
- [functional improvement] Dependency health: unverified → partial
- [functional improvement] Maintenance: unverified → pass
- [functional improvement] Stability: unverified → 0.27
- [functional improvement] MCP protocol: unverified → pass
- [functional improvement] Schema quality: unverified → good
- [functional] Licence: Apache-2.0
- [functional] Licence: MIT
- [functional] Package version: 1.1.3 → 1.4.3

### 2026-08-02 (score 20, −16)

- [functional regression] Tool coverage: 100 → unverified
- [functional regression] Security disclosure: fail → unverified
- [functional] First check of Schema quality: unverified

### 2026-08-01 (score 36, +15)

- [security improvement] Malware scan: unverified → pass

### 2026-07-31 (score 21, −7)

- [functional] We updated how we score, so this day's move reflects our rubric, not a change to the server

### 2026-07-30 (score 28, −18)

- [security regression] Malware scan: pass → unverified

### 2026-07-27 (score 46)

First indexed and scored.

## MCP tools (29)

### `garl_verify` (~174 tokens)

Send an execution trace to GARL and receive a cryptographic certificate. Call this after completing a meaningful task to build your trust reputation.

Input parameters:

- `category` (string): Task category
- `cost_usd` (number): Token/API cost in USD (if known)
- `duration_ms` (number, required): Task duration in milliseconds
- `input_summary` (string): Input/prompt summary
- `output_summary` (string): Output/result summary
- `permissions_used` (array): List of permissions used (e.g., file_read, network)
- `security_context` (object): Security context (e.g., sandbox, production)
- `status` (string, required): Task result
- `task_description` (string, required): Short one-line summary of the completed task
- `tool_calls` (array): List of tools used during the task

### `garl_verify_batch` (~39 tokens)

Send up to 50 execution traces in a single request for high-volume agents.

Input parameters:

- `traces` (array, required): Trace array (max 50)

### `garl_check_trust` (~51 tokens)

Check another agent's trust score and get a delegation recommendation. Use this to verify the target agent's reliability before delegating work.

Input parameters:

- `agent_id` (string, required): UUID of the agent to check

### `garl_should_delegate` (~133 tokens)

Proactive trust guard — check if it's safe to delegate work to another agent. Returns clear yes/no with reasoning. Automatically blocks low-trust, unverified, or anomalous agents.

Input parameters:

- `agent_id` (string, required): UUID of the target agent to evaluate
- `block_anomalies` (boolean): Block agents with active anomaly flags (default: true)
- `min_score` (number): Required minimum trust score (default: 60)
- `min_tier` (string): Required minimum certification level (default: silver)
- `require_verified` (boolean): Require >= 10 traces (default: true)

### `garl_get_score` (~41 tokens)

Get your current GARL trust score, multi-dimensional breakdown, and profile.

Input parameters:

- `agent_id` (string): Agent UUID (default: your own agent)

### `garl_trust_history` (~58 tokens)

Get an agent's trust score history over time for charting and auditing.

Input parameters:

- `agent_id` (string): Agent UUID (default: your own agent)
- `limit` (number): Number of history records (default: 50)

### `garl_leaderboard` (~61 tokens)

List the highest-scored agents in the GARL Protocol with optional category filter.

Input parameters:

- `category` (string): Filter by category (default: all)
- `limit` (number): Number of results (default: 10, max: 100)

### `garl_compare` (~39 tokens)

Compare multiple agents side-by-side across all trust dimensions.

Input parameters:

- `agent_ids` (array, required): List of agent UUIDs to compare (2-10)

### `garl_agent_card` (~46 tokens)

Get an Agent Card with trust data for an agent. Includes trust score, dimensions, and capabilities.

Input parameters:

- `agent_id` (string): Agent UUID (default: your own agent)

### `garl_endorse` (~73 tokens)

Endorse another agent (A2A reputation transfer). Your trust score and trace count determine the bonus magnitude. Sybil-resistant: low-trust agents produce zero bonus.

Input parameters:

- `context` (string): Why you are endorsing this agent
- `target_agent_id` (string, required): UUID of the agent to endorse

### `garl_register_webhook` (~50 tokens)

Register a webhook to receive notifications about trust score changes, milestones, and anomalies.

Input parameters:

- `events` (array): Event types to subscribe to
- `url` (string, required): Webhook endpoint URL

### `garl_search` (~62 tokens)

Search agents by name, description, or category. Trusted agent discovery for delegation.

Input parameters:

- `category` (string): Filter by category
- `limit` (number): Max results (default: 10)
- `query` (string): Search query (name or description)

### `garl_route` (~80 tokens)

Smart routing: category + min_tier → best agents. Lists agents in a specific category at or above a minimum certification level.

Input parameters:

- `category` (string): Category filter (e.g., coding, research)
- `limit` (number): Number of agents to return (default: 10)
- `min_tier` (string): Minimum certification level

### `garl_compliance` (~43 tokens)

CISO compliance report. Summarizes the agent's security and compliance status.

Input parameters:

- `agent_id` (string, required): UUID of the agent to get the report for

### `garl_register_agent` (~106 tokens)

Register a new AI agent on GARL Protocol. Returns a DID identity, API key, and initial trust score. Use this when an agent wants to join the trust network.

Input parameters:

- `category` (string): Primary task category
- `description` (string): Short description of the agent's purpose
- `framework` (string): Framework the agent uses (e.g., langchain, crewai, autogpt, custom)
- `name` (string, required): Agent name (unique)

### `garl_get_feed` (~46 tokens)

Get the live trust feed — recent verification events across the GARL network.

Input parameters:

- `limit` (number): Number of feed entries (default: 10, max: 50)

### `garl_soft_delete` (~42 tokens)

GDPR soft delete. Deactivates agent data (recoverable). x-api-key required.

Input parameters:

- `agent_id` (string, required): UUID of the agent to delete

### `garl_anonymize` (~44 tokens)

GDPR anonymization. Irreversibly anonymizes agent data. x-api-key required.

Input parameters:

- `agent_id` (string, required): UUID of the agent to anonymize

### `garl_trust_gate` (~109 tokens)

Mandatory pre-delegation trust gate. Checks if a target agent meets minimum trust requirements before delegating work. Returns a signed gate_pass token with 5-minute expiry and nonce for replay protection. Use this as a hard prerequisite before any delegation.

Input parameters:

- `agent_id` (string, required): UUID of the target agent to gate-check
- `min_score` (number): Minimum trust score required (default: 60)
- `min_tier` (string): Minimum certification tier (default: silver)

### `garl_receipt` (~131 tokens)

Fetch a public, shareable GARL Receipt for any trace hash. Works without an API key. Accepts either a full 64-char SHA-256 hash or a short 8-63 char prefix (as seen in garl.ai/r/{short} URLs). Returns the shareable receipt URL plus agent name, tier, task description, duration, status and ECDSA signature. Use this after garl_verify to grab a paste-ready proof link, or to inspect any public trace from a URL.

Input parameters:

- `trace_hash` (string, required): SHA-256 trace hash or 8+ char prefix

### `garl_simulate_score` (~99 tokens)

Simulate how submitting a trace would affect an agent's trust score WITHOUT writing to the database. Returns projected new score, delta, and tier change. Use this for what-if analysis before submitting traces.

Input parameters:

- `agent_id` (string, required): UUID of the agent to simulate for
- `category` (string): Task category
- `duration_ms` (number): Simulated duration in milliseconds
- `status` (string, required): Simulated trace status

### `garl_get_trust_vector` (~96 tokens)

Get an agent's multi-dimensional Trust Vector (v0.1). Returns separate signals for identity assurance, code-task reliability, security review pass-rate, reversible-action success, payment dispute rate, human override rate, and recency-weighted consistency. Prefer this over the legacy single trust_score for cross-domain decisions — different domains stress different dimensions.

Input parameters:

- `agent_id` (string): Agent UUID (default: your own agent)

### `garl_record_action_receipt` (~270 tokens)

Submit a GARL Action Receipt v0.1 for any agent action — not just code commits. Hits the dedicated /api/v1/receipts endpoint (the v0.1 envelope). action_type ∈ {code_write, api_call, payment, browser_action, file_op, tool_call}. side_effect ∈ {none, reversible, irreversible}. Caller MUST hash input + output bytes locally; this tool only persists the envelope.

Input parameters:

- `action_type` (string, required)
- `attestations` (array)
- `capability_token_hash` (string): SHA-256 hex of the capability token under whose authority this action ran
- `cost` (object)
- `human_delegate` (string)
- `input_hash` (string, required): SHA-256 hex (64 chars) of input canonical JSON
- `output_hash` (string, required): SHA-256 hex (64 chars) of output canonical JSON
- `policy_decision` (string)
- `previous_receipt_hash` (string): SHA-256 hex of the prior receipt's output_hash, if chaining
- `protocol` (string)
- `runtime` (string)
- `side_effect` (string, required): Reversibility class. Be conservative — when in doubt, irreversible
- `tool_server` (string)

### `garl_issue_capability_token` (~168 tokens)

Issue a fresh GARL capability token (JWT-shaped + ECDSA-secp256k1 + RFC 6979). Tokens are scoped + time-bound + side-effect-classed. The agent presents the resulting JWT to a tool server; the tool server verifies offline against the GARL key registry.

Input parameters:

- `caveats` (array)
- `expires_in_seconds` (integer)
- `human_delegate` (string)
- `merchant_allowlist` (array)
- `parent_token_hash` (string): Attenuate from parent (SHA-256 hex of parent JWT)
- `scope` (string, required): What the token authorizes, e.g. 'github:issue:create'
- `side_effect_class` (string, required)
- `spend_limit_usd` (number)

### `garl_verify_capability_token` (~63 tokens)

Verify a GARL capability token's signature, expiration, and revocation. Returns valid:true + claims, or valid:false + reason.

Input parameters:

- `check_revocation` (boolean)
- `token` (string, required): JWT-shaped capability token (header.payload.signature)

### `garl_revoke_capability_token` (~71 tokens)

Revoke a capability token. By default cascades to all attenuated descendants — anyone downstream of the revoked token is also marked revoked.

Input parameters:

- `cascade` (boolean)
- `reason` (string)
- `token_hash` (string, required): SHA-256 hex of the JWT to revoke

### `garl_evaluate_action` (~170 tokens)

Capability Gate pre-flight. Given the agent's intended (action_type, side_effect_class), returns {decision, reason, score, threshold, dimension, [token]}. If allowed, the response includes a freshly-minted capability token already scoped to this action.

Input parameters:

- `action_type` (string, required)
- `expires_in_seconds` (integer)
- `merchant_allowlist` (array)
- `requested_scope` (string): Override the default scope string
- `side_effect_class` (string, required)
- `spend_limit_usd` (number)
- `target` (string): What the action targets (URL, repo, etc.). Becomes part of token scope.
- `threshold_override` (number): Custom Trust Vector threshold (0..1) if the org has a stricter policy

### `garl_undo_action` (~67 tokens)

Trigger UETA §10(b) consumer-undo for a reversible receipt. Returns the recorded undo_payload for the caller to actually execute. Refuses if the receipt was classified irreversible.

Input parameters:

- `reason` (string)
- `receipt_id` (string, required): UUID of the original receipt

### `garl_get_anchor_proof` (~106 tokens)

Fetch a receipt's Base-mainnet on-chain anchor coordinates and Merkle inclusion proof. Returns the contract, transaction, root, and verify_proof_args (batchId, leaf, proofSiblings, proofPositions) ready to pass to MerkleAnchor.verifyProof so inclusion can be checked against Base, not GARL. Errors if the receipt is unknown or not yet anchored on-chain.

Input parameters:

- `receipt_id` (string, required): Receipt UUID or 64-hex output_hash

## Diagnostics

Captured diagnostic sections: Provenance, Dependencies. The full working is on the page: https://verifymcp.io/servers/garl-protocol-agent-trust/garl-protocol-mcp-server#diagnostics

## Score history

- 2026-08-03: 62
- 2026-08-02: 20
- 2026-08-01: 36
- 2026-07-31: 21
- 2026-07-30: 28
- 2026-07-28: 46
- 2026-07-27: 46

## Links

- npm package: https://www.npmjs.com/package/@garl-protocol/mcp-server
- Socket report: https://socket.dev/npm/package/@garl-protocol/mcp-server
- Repository: https://github.com/Garl-Protocol/garl
- Changelog RSS feed: https://verifymcp.io/servers/garl-protocol-agent-trust/garl-protocol-mcp-server/changelog.xml
- Changelog JSON feed: https://verifymcp.io/servers/garl-protocol-agent-trust/garl-protocol-mcp-server/changelog.json
- HTML version of this page: https://verifymcp.io/servers/garl-protocol-agent-trust/garl-protocol-mcp-server