Skip to content
PipMCP

Logs

Execution logs, redaction, the audit log, usage metering and retention.

PipMCP records every tool call and every configuration change. Secrets are removed before anything is stored.

Execution log

Open Executions to see every tool call, newest first. Filter by status, for example Pending approval or Failed.

Each execution records:

  • Tool, integration, endpoint and the token used.
  • Client name and version, and the protocol version.
  • Status: succeeded, failed, denied, confirmation_required, cancelled or running, plus the approval status where relevant.
  • Risk level and the policy decision with its reason (which preset or policy decided, and how a confirmation was given).
  • Input, redacted.
  • Result summary: shape, item count, top-level keys and a short redacted preview, or the normalized error code and message.
  • Upstream steps: each HTTP request made, with method, status and timing.
  • Duration and a trace ID for support requests.
  • Whether the call was billable.

The same data is available through the API: GET /api/v1/executions and GET /api/v1/executions/{id}. See Management API.

Redaction

Before inputs, outputs, errors or request details are stored, they pass through a redactor:

  • By field name: values of fields and headers such as authorization, cookie, x-api-key, api_key, access_token, refresh_token, client_secret, password, secret, token, private_key and signature are replaced with [REDACTED].
  • By value: strings that look like credentials are masked wherever they appear, including Bearer … and Basic … values, JWTs, Stripe-style keys, GitHub tokens, AWS access key IDs, Slack tokens and PipMCP's own pmcp_ and pmk_ tokens.
  • Known secrets: the exact secret values of the connection used for a call are masked in anything logged for that call.

Redaction happens before storage, so full values are never written to logs.

Audit log

Configuration changes are recorded separately in the audit log (/app/audit), for example:

  • integrations imported, published or deleted; tools edited, published or live-tested;
  • connections created, updated, tested or deleted;
  • endpoints created or changed, and tokens created or revoked;
  • policies changed, approvals given, members invited or removed;
  • management API tokens created or revoked.

Each entry records who did it (a user, the API on a user's behalf, or an MCP client), when, and what changed. Secrets never appear in the audit log.

Usage

Usage shows agent actions for the current calendar month against your plan's limit, with a breakdown by tool, client and status. An agent action is one tools/call that PipMCP executes against your API. Not counted:

  • tools/list and other discovery requests,
  • calls denied by policy or waiting for confirmation,
  • calls rejected for invalid input,
  • live tests from the dashboard and calls to the demo API.

GET /api/v1/usage returns the same numbers.

Retention

Execution logs are deleted automatically once they are older than your plan's retention period (for example 30 days on Developer, 90 days on Pro, 365 days on Business; custom on Enterprise). Executions that went through the approval flow are kept. Pruning runs daily.