Pipelock includes an operator console: a read-only view over the signed evidence Pipelock records for the traffic and tool calls it mediates. It shows what your agents did, lets you inspect and verify the receipt behind each decision, and gives operators read-only surfaces for exemptions, budgets, keys, and fleet state (mutations happen through the CLI and Conductor commands, not the console). It is deliberately honest about the limits of what it can prove.
What you are looking at
The console filters by authorization: a Pro license opens the agent and evidence surfaces; Enterprise adds the live fleet, workbench, and incident surfaces. Unconnected sources render explicit empty states rather than implying health.
Overview (Enterprise view, live fleet posture populated). It leads with ‘what can I prove?’ and shows honest empty states, like an unconfigured conductor source, instead of faking a healthy dashboard; on a Pro license the fleet section shows an Enterprise-required or no-source state.
Evidence (Pro). Each session’s evidence is scored on four separate axes: Authentic (signed by the expected key), Untampered (hash chain intact), Anchored (whether it was externally anchored), and Completeness (LIMITED, never COMPLETE). The console never collapses these into a single ‘verified.’
Fleet (Enterprise). Enrolled followers report runtime, signed-audit, applied-state, and drift status; stale or diverged followers render as non-green states rather than being assumed healthy.
Free, Pro, and Enterprise
| Capability | Free | Pro (agents) | Enterprise (fleet) |
|---|---|---|---|
| Inspect one recorded session (read-only) | yes | yes | yes |
| Verify signed evidence + coverage certs offline | yes | yes | yes |
| Operator console: Overview, Evidence, Exemptions, Agents, Budgets, Trust & Keys | no | yes | yes |
| Generate coverage certificates | no | yes | yes |
| Live Fleet, Workbench, Incident surfaces | no | no | yes |
| SIEM event forwarding | no | no | yes |
Single-session evidence viewing and offline verification are free. The console itself is a Pro and Enterprise capability.
The evidence is honest about what it proves
The scorecard keeps four claims separate on purpose, because collapsing them into one word is how “audit logs” oversell themselves:
- Authentic — the record was signed by the expected key. Verifying offline against the expected pinned public key is what makes this meaningful; a valid signature alone does not establish a trusted signer.
- Untampered — the loaded hash chain is intact, so loaded records were not altered or reordered and no sequence gap was found. Tail truncation, omitted sources, and direct egress stay completeness and anchoring questions.
- Anchored — whether the record was additionally committed to an external log. The scorecard reports Not anchored unless an external inclusion proof is verified, and shows that default state rather than folding it into Authentic or Untampered.
- Completeness — reported as LIMITED, never COMPLETE. It covers the sessions and window Pipelock mediated, with gaps and standing exclusions visible.
That last point is the boundary that matters: Pipelock proves what it mediated, not what it never saw. If you need every path forced through the proxy, that is a deployment control (see the FAQ), not a claim the binary makes on its own.
Run it yourself
You do not need the console, or a license, to see the proof model:
pipelock demo --receipts-dir /tmp/r
receipt="$(find /tmp/r -maxdepth 1 -name '*.json' | sort | head -n 1)"
pipelock verify-receipt "$receipt" --key "$(cat /tmp/r/signer.pub)"
That throws real attacks at a local instance, blocks them, and signs a receipt for each; you then verify one offline against the published key. Without --key, verification is structural only and reports UNPINNED. To view a recorded session as a read-only evidence page, free:
pipelock evidence serve --receipt-dir /var/lib/pipelock/evidence --session agent-a
Or try the browser challenge: a real agent guards a planted secret, and you try to get it out.
Upgrading
Pipelock 3.2 is the current hardening line with fail-closed changes to receipt verification, reload behavior, and signing-key requirements. Read the GitHub releases before upgrading.