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 viewrenders 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.
1 maintainer
Release files for ponens 1.15.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| ponens-1.15.3.tar.gz | 328.7 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| ponens-1.15.3-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 679.5 kB
Release files / ponens-1.15.3.tar.gz
| Download URL | ponens-1.15.3.tar.gz |
|---|---|
| Size | 328.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
a73c825f83883c7ac4db2ea8f92422a4e6bd898a518fc68e181c6723324ff7d7
|
|
BLAKE2b-256 checksum How to use checksums |
b7359651617ae60779a93d68219401ae2f3405f531f750c7e6c36e6dec5f45c9
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.6.11
|
Release files / ponens-1.15.3-py3-none-any.whl
| Download URL | ponens-1.15.3-py3-none-any.whl |
|---|---|
| Size | 350.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
66dbb6ca4aab25b40d09ecc7df1715cf12166f1f4baf5d57e6b140e30209d9fd
|
|
BLAKE2b-256 checksum How to use checksums |
edf5f28e1738a675d95bad76d5f68351cb2f9f8640517f2e86c19412a4b76dbd
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.6.11
|
