ponens

Turn an AI agent's session into a reviewable, verifiable reasoning record — and govern it.

pip install ponens

When an AI coding agent finishes, all you really have is a diff and a chat log. ponens turns that into a reasoning trace: a structured, reviewable record of what was built and why — the steps, the decisions, the artifacts and their lineage, the declared goal and whether it was met, what was verified, and what was left uncertain — at the altitude a reviewer actually reads, not a replay of the dialogue.

Reasoner-agnostic, local-first (no backend required), MIT-licensed.

Website & policy gallery: https://ponens.dev

Quick start

pip install ponens

ponens emit -o trace.json          # capture an agent session as a reasoning trace
ponens trace view trace.json       # open the interactive viewer (flow · lineage · goals · policies)
ponens trace grade trace.json      # score the record on a reviewer-oriented rubric
ponens trace check trace.json      # computable governance — machine-checked policies, PASS/FAIL

Already instrumented with OpenTelemetry or Langfuse? Bring those traces in and add the governance layer:

ponens otel import spans.json -o trace.json
ponens langfuse import lf.json  -o trace.json

What you get

• A reasoning trace, not a transcript — atomic actions, typed artifacts, and a lineage DAG (what produced what), plus a declared residual surface (the gaps the agent left open).
• Goals graded met vs certified — did it meet its definition of done (resolved from the trace's own evidence), and did a reviewer other than the doer confirm that was the right definition? "Done" separated from "right."
• Computable governance — organizational and best-practice policies as machine-checked formulas over the trace (temporal + structural), a deterministic PASS/FAIL gate that runs offline in CI. Ready policy packs (formal-methods, the FIX Community's AI guidelines, and more) in the gallery.
• An interactive viewer — ponens trace view renders the flow, the lineage graph, the goals, and the policy verdicts in a single self-contained page.

Learn more

• Website & policy gallery — https://ponens.dev
• Specs (trace, policy language, goal faithfulness) — https://ponens.dev/spec
• Source & issues — https://github.com/imandra-ai/ponens

License

MIT — © Imandra, Inc.