Audit a whole MCP client config: one PASS / REVIEW / BLOCK / UNKNOWN per entry

Resolves every entry of a config document without guessing (url → remote_url, npx → npm package, uvx → PyPI package, canonical alias → server_name) and evaluates each resolved server under one built-in policy through the same core as /v1/preflight. Returns CENSUS-AUDIT/1 JSON and text, valid_until_epoch for shell caches and an exit code for CI. UNKNOWN is a real verdict and is never upgraded to PASS. Keyed: 1 credit per config digest per UTC day. Keyless: one config of ≤25 entries per IP per UTC day, repeats free. Not a malware scan; PASS is not a sandbox. Contract: https://mcpcensus.com/agent-setup/audit.md

post/v1/audit

Request

  • Base URL: https://api.mcpcensus.com
  • URL: https://api.mcpcensus.com/v1/audit
  • Auth: one of:
    • none
    • API key in header x-api-key

Request body

format'json' | 'toml'
policy_id'builtin:baseline' | 'builtin:first-party' | 'builtin:strict'
previous_tools_digestsobject

alias → tools_digest from your last audit; sets tools_drift per entry

strictboolean

exit_code 2 when any entry is REVIEW or UNKNOWN

mcpServersobject
serversobject
mcp_serversobject

Response

Completed audit

okunknown
spec'CENSUS-AUDIT/1'
as_ofstring date-time
valid_untilstring date-time nullable

Minimum over resolved entries

valid_until_epochinteger nullable
config_digeststring
config_formatstring
decision_scope'public_evidence_policy'
exit_code0 | 1 | 2

1 on any BLOCK; 2 with strict on REVIEW/UNKNOWN; else 0. Mirrored in the x-census-exit-code header

textstring

The CENSUS-AUDIT/1 text block (also returned alone for Accept: text/plain)

limitationsstring[]
meterobject

mode credits (cost, repeat, remaining) or keyless (configs_today, limit)

audit_run_idstring nullable

Changes

No recorded changes to this operation across all 1 revision of this API.