Product MCP

Org-scoped Intertrace tools over Streamable HTTP at /mcp. Clerk authenticates humans; Intertrace issues MCP tokens.

Two MCP URLs

Product MCP (https://platform.intertrace.ai/mcp) is Intertrace as a server: findings, assets, authorize, HITL. Identity is Clerk for the consent screen; the MCP client then uses an Intertrace token (itm_…), not a Clerk access token.

Gateway MCP proxy (https://intertrace.fly.dev/mcp) is how you send your agents through Intertrace to customer MCP servers. See MCP servers.

Connect

Remote MCP URL: https://platform.intertrace.ai/mcp. OAuth 2.1 with PKCE S256. Resource indicator: the same URL. Scopes: mcp:read, mcp:write, mcp:authorize, offline_access.

  1. 1

    Claude.ai

    Settings → Connectors → Add custom connector named Intertrace. URL https://platform.intertrace.ai/mcp. Sign in with Clerk; Intertrace issues itm_… tokens.

  2. 2

    ChatGPT

    Settings → Apps & Connectors → add a custom MCP connector. Same URL. Do not paste a Clerk or OpenAI JWT.

  3. 3

    Cursor / Claude Code

    Add the snippet below. OAuth is discovered from the 401 WWW-Authenticate challenge.

  4. 4

    Devin

    Settings → Connections → MCP servers → Add a custom MCP. Transport HTTP, URL https://platform.intertrace.ai/mcp, OAuth. Callback https://api.devin.ai/mcp/oauth/callback is already allowed. Walkthrough: Devin.

{
  "mcpServers": {
    "intertrace": {
      "url": "https://platform.intertrace.ai/mcp"
    }
  }
}

Cursor and Claude Code discover OAuth from the 401 WWW-Authenticate challenge. API keys work as Authorization: Bearer on the same URL.

curl -sS -D - https://platform.intertrace.ai/mcp \
  -H "Authorization: Bearer itm_…" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"docs","version":"1"}}}'

curl -sS https://platform.intertrace.ai/mcp \
  -H "Authorization: Bearer itm_…" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}'

Revoke clients in Settings → MCP. Calls share the org monthly request quota (gateway_requests).

Tools

Every tool is org-scoped. Auth is headers only — never put tokens in arguments. Writes need mcp:write; authorize needs mcp:authorize.

  • intertrace_health Intertrace health — Public Intertrace control-plane health (database, ingest, runtime). No tenant data.(mcp:read)
  • intertrace_whoami Who am I — Return the authenticated organization, role, scopes, and auth source for this MCP session.(mcp:read)
  • intertrace_list_traces List traces — List recent Intertrace telemetry traces for the authenticated organization.(mcp:read)
  • intertrace_get_trace Get trace — Fetch one telemetry trace by id if it belongs to the authenticated organization.(mcp:read)
  • intertrace_list_findings List findings — List Intertrace findings with optional status, source, and severity filters.(mcp:read)
  • intertrace_get_finding Get finding — Fetch one finding with prompt fields revealed according to the caller's role.(mcp:read)
  • intertrace_list_assets List assets — List AI assets in the authenticated organization (secrets redacted).(mcp:read)
  • intertrace_get_asset Get asset — Fetch one AI asset. Provider API keys are never returned in plaintext.(mcp:read)
  • intertrace_discovery_coverage Discovery coverage — Connector fidelity coverage: configured vs fixture vs live, plus estate gaps.(mcp:read)
  • intertrace_list_gateway_keys List gateway ingest keys — List gateway ingest keys (key id / prefix only). Never returns signing secrets.(mcp:read)
  • intertrace_list_mcp_inventory List MCP inventory — List customer MCP servers Intertrace governs (upstream URLs, not vault credentials).(mcp:read)
  • intertrace_list_runtime_sessions List runtime sessions — List runtime sessions with risk/alignment summaries.(mcp:read)
  • intertrace_get_runtime_session Get runtime session — Fetch one runtime session story and related evidence for the authenticated org.(mcp:read)
  • intertrace_list_runtime_decisions List runtime decisions — List recent runtime verification / authorize decisions for the organization.(mcp:read)
  • intertrace_list_usage List usage — Return organization usage counters for the current billing period.(mcp:read)
  • intertrace_search Search — Org-scoped search across findings, assets, and traces. Not a staff/ops search.(mcp:read)
  • intertrace_posture_evaluate Evaluate posture — Evaluate posture checks for org assets. dry_run=true (default) does not persist findings.(mcp:read)
  • intertrace_policy_simulate Simulate policy — Dry-run events through local authorize.rego-priority rules. Does not enforce live traffic.(mcp:read)
  • intertrace_list_assessments List assessments — List recent assessment runs for the organization.(mcp:read)
  • intertrace_industry_readiness Industry solution readiness — Computed industry-solution readiness. Catalog wording only — not a compliance certificate.(mcp:read)
  • intertrace_authorize_action Authorize action — Ask Intertrace whether an intended agent action is allowed. Uses the same policy simulator as /api/policy/simulate and persists an authorize outcome. Live Fly gateway enforcement remains the data-plane kernel.(mcp:authorize)
  • intertrace_update_finding_status Update finding status — Set a finding status (in_progress, resolved, accepted, false_positive) with the same audit trail as the dashboard.(mcp:write)
  • intertrace_hitl_approve Approve HITL request — Approve a pending runtime approval request for this organization.(mcp:write)
  • intertrace_hitl_deny Deny HITL request — Deny a pending runtime approval request for this organization.(mcp:write)
  • intertrace_protect_asset Protect asset — Mark an asset as protected (counts against the protected-agent quota).(mcp:write)
  • intertrace_start_assessment Start assessment — Queue an assessment run for an asset using an existing probe profile.(mcp:write)
  • intertrace_connection_opportunities Connection opportunities — Deterministic shared enforcement chokepoints: how many agents each infrastructure connection would cover. Does not deploy.(mcp:read)
  • intertrace_protect_estate_plan Protect estate plan — Generate the Protect My AI Estate connection plan (ranked chokepoints, coverage, bypasses). Does not apply infrastructure.(mcp:read)