agentscore-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| GITHUB_TOKEN | No | GitHub personal access token (optional, increases rate limit to 5,000/hr). | |
| MOLTBOOK_API_KEY | No | Required for Moltbook adapter. | |
| AGENTSCORE_ADAPTER | No | Choose the data source adapter: demo, github, json, or moltbook. | demo |
| AGENTSCORE_ENFORCE | No | If true, policy gate can block risky results. | false |
| AGENTSCORE_AUDIT_LOG | No | Set false to suppress structured policy audit events. | auto |
| AGENTSCORE_CACHE_TTL | No | Score cache TTL in seconds. | 86400 |
| AGENTSCORE_DATA_PATH | No | Required for JSON adapter. Path to JSON data file. | |
| AGENTSCORE_HTTP_HOST | No | Bind host for HTTP transport. | 127.0.0.1 |
| AGENTSCORE_HTTP_PATH | No | MCP endpoint path for HTTP transport. | /mcp |
| AGENTSCORE_HTTP_PORT | No | Bind port for HTTP transport. | 8787 |
| AGENTSCORE_TRANSPORT | No | Transport mode: stdio or http (Streamable HTTP server mode). | stdio |
| AGENTSCORE_AUDIT_TOKEN | No | Optional bearer token required for policy/audit endpoints. | |
| AGENTSCORE_PUBLIC_MODE | No | If true, requires explicit adapter and blocks demo mode. | false |
| AGENTSCORE_ENABLED_TOOLS | No | Comma-separated tool allow-list (agentscore, sweep, xray). | agentscore,sweep,xray |
| AGENTSCORE_RATE_LIMIT_MS | No | Moltbook adapter request delay in milliseconds. | 200 |
| AGENTSCORE_HTTP_AUTH_TOKEN | No | Optional bearer token required for /mcp HTTP endpoint. | |
| AGENTSCORE_POLICY_MIN_SCORE | No | Minimum allowed score when policy is enforced. | 550 |
| AGENTSCORE_AUDIT_MAX_ENTRIES | No | In-memory cap for retained policy audit events. | 500 |
| AGENTSCORE_POLICY_BLOCK_FLAGS | No | Comma-separated flag substrings that trigger blocking. | prompt injection,manipulation keyword,account not claimed |
| AGENTSCORE_POLICY_FAIL_ON_ERRORS | No | If true, any per-handle scoring errors trigger blocking. | false |
| AGENTSCORE_POLICY_TRUSTED_ADAPTERS | No | Comma-separated adapters allowed in enforced mode. | github,json,moltbook |
| AGENTSCORE_POLICY_BLOCK_THREAT_LEVELS | No | Comma-separated blocked sweep levels (SUSPICIOUS, COMPROMISED). | COMPROMISED |
| AGENTSCORE_POLICY_BLOCK_RECOMMENDATIONS | No | Comma-separated blocked recommendations (TRUST, CAUTION, AVOID). | AVOID |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| agentscoreA | Score 1-5 AI agents for trustworthiness. Returns investigation briefings, comparison verdicts, risk flags, and trust badges. Use for: trust checks, comparisons, verifications, badge generation. |
| sweepA | Analyze a thread or conversation for manipulation patterns. Detects coordinated bots, sock puppets, and astroturfing campaigns by scoring every participant and checking for content similarity, timing anomalies, and amplification signals. |
| xrayA | X-ray content for hidden AI-targeted payloads. Detects concealed instructions in markdown, HTML, code, and text before an agent consumes it, including hidden comments, invisible unicode, CSS-hidden text, encoded payloads, code comments, and structural hiding tricks. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 3 tools
Each tool targets a different artifact: agentscore evaluates AI agents, sweep analyzes conversation participants, and xray inspects static content for hidden payloads. There is slight overlap in the language of scoring and risk flags between agentscore and sweep, but the descriptions make the intended use clear.
All tool names are lowercase single-word names, giving the set a consistent stylistic pattern. However, they do not follow a descriptive verb_noun convention, and agentscore is a compound while sweep and xray are metaphorical, so function is not immediately inferable from naming alone.
Three tools is a well-scoped size for a focused trust-safety investigation server. Each tool covers a distinct and meaningful capability, and none feels redundant or extraneous.
The tool surface covers the core trust-safety workflow: agent trust scoring, conversation manipulation analysis, and hidden content payload detection. Minor gaps exist around persistence, historical investigation lookup, or more granular export/reporting, but these are workable limitations rather than blocking dead ends.