For teams using external agents and MCP servers: inspect measured evidence, watch changes, and keep a record of what your team accepted. Unknown evidence stays explicit.
Have a client configuration? Extract and check its public endpoints locally. Maintaining integrations for clients? Explore the agency pilot.
Runtime observation: 2026-10-29T03:05:37.422820+00:00 · State: live
Watch availability, tools and parameter schemas, authentication access, endpoint changes and lifecycle notices. Only complete observed schemas are compared. OAuth scope and consent-screen changes are not currently measured.
Sign in to watch this MCP serverSign-in required. Agent and MCP watches share the existing Free / Pro / Enterprise budget. Manage watches.
{
"auth": "anonymous",
"availability": "live",
"endpoint": "https://completionkit.com/mcp",
"legal_identity": null,
"lifecycle": null,
"observed_at": "2026-10-29T03:05:37.422820+00:00",
"registry_status": null,
"subject": {
"id": "com.completionkit/evals",
"kind": "mcp"
},
"tools": {
"names": [
"agreements_create",
"agreements_list",
"datasets_create",
"datasets_create_from_url",
"datasets_delete",
"datasets_get",
"datasets_list",
"datasets_update",
"judges_compare",
"judges_replay",
"metric_groups_create",
"metric_groups_delete",
"metric_groups_get",
"metric_groups_list",
"metric_groups_update",
"metric_versions_dismiss",
"metric_versions_list",
"metric_versions_publish",
"metrics_create",
"metrics_delete",
"metrics_get",
"metrics_list",
"metrics_suggest_variants",
"metrics_update",
"promptfoo_import",
"prompts_create",
"prompts_delete",
"prompts_get",
"prompts_list",
"prompts_publish",
"prompts_suggest_improvement",
"prompts_update",
"provider_credentials_create",
"provider_credentials_delete",
"provider_credentials_get",
"provider_credentials_list",
"provider_credentials_update",
"responses_get",
"responses_list",
"runs_create",
"runs_delete",
"runs_generate",
"runs_get",
"runs_list",
"runs_regrade",
"runs_rerun",
"runs_retry_failures",
"runs_update",
"tags_create",
"tags_delete",
"tags_get",
"tags_list",
"tags_update",
"usage_get"
],
"names_truncated": false,
"sha256": "37766d4e223766f6ad95020eff5459075ed9b0904f581154102e54f6f42ae532"
},
"tools_complete": true,
"tools_observed_at": "2026-10-29T03:05:37.422820+00:00",
"version": null
}Keep an accepted snapshot in agents.lock.json. CI compares it with current evidence and returns allow, review or block with reasons, before/after data and policy hashes.
Default gaps require review. Configure stricter handling explicitly. A lock records your acceptance; it is not a supplier attestation.
API referenceVerified domain owners can publish up to five exact-match sandbox fixtures and opt in to a test run. Results include fixture version, latency, output hashes and comparable regressions. Cost remains unknown when unobserved.
Run with POST /api/v1/dependencies/task-checks?domain=… while signed in. Read the schema and example in the integration guide.
Domain control, a declared business identifier and an active name-matched company record are separate evidence. Zero verified records means zero recorded proof.
Current coverage and measurement date · Verify your domainVerify issuer signatures, subject and credential freshness using the public key document. Keep its task evidence categories separate from Agenstry measurements.
A valid SEAL does not establish domain ownership, legal identity, key possession or execution permission.
Credential verification API · Issuer documentationSmithery endpoint migration · Agent monitoring · Methodology