Pawl is a Python-based framework for registry-driven verification of claims and autonomous research workflows, focusing on per-claim gates, a memory-backed registry, and an autonomous loop to promote verified progress.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (2 so far).
What it is
Pawl is a Registry-as-ground-truth verification and autonomous research framework for agent-driven development. It uses a registry (registry.csv) to record claims, gates, and outcomes, with an emphasis on honest negatives and frozen gates prior to implementation. It operates as a zero-dependency, pure-Python approach that can be adopted by copying a folder and writing a small config file.
How it works
Pawl maintains a memory of attempts and outcomes, gates that cannot be breached unless verified, and a loop for running experiments unattended without corrupting the record. The repository includes a runner that promotes claims to Verified or marks them as Failed, and a memory/indexing system (BM25) to query fronts, claims, and plans. The structure includes tools such as init_project.py, register.py, init_folder.py, run.py, scan.py, memory.py, and a dedicated autoresearch layer to enable autonomous experiments. The system enforces gates defined a priori and records Honest Negatives as progress.
Getting started
Quickstart guidance (as described in the README)
- Copy the pawl folder into your repo (or keep it as a sibling checkout). TOOLS below = path to pawl/tools.
-
- Bootstrap your project (creates verify.json, registry.csv, tracks.json): cd ~/my-project python3 $TOOLS/init_project.py my-project --plans-dir docs/plans
-
- Register your first claim — a falsifiable sentence with thresholds:
python3 $TOOLS/register.py VERIFY.CORE.CSV_PARSER.01
"CSV parser handles quoted fields and CRLF; 12 fixture files round-trip \byte-identical; parse of 100k rows <= 2 s on the fixed bench"
- Register your first claim — a falsifiable sentence with thresholds:
python3 $TOOLS/register.py VERIFY.CORE.CSV_PARSER.01
-
- Create its folder (README + run.py skeleton): python3 $TOOLS/init_folder.py VERIFY.CORE.CSV_PARSER.01
-
- Freeze the gates in the generated run.py, THEN implement the feature.
-
- Official pass — writes Verified/Failed back to the registry: python3 $TOOLS/run.py VERIFY.CORE.CSV_PARSER.01
-
- Query your project's memory: python3 $TOOLS/memory.py fronts python3 $TOOLS/memory.py query "csv parser" python3 $TOOLS/memory.py show VERIFY.CORE.CSV_PARSER.01
A fully executed example lives in examples/demo-project/ — including one Verified claim and one honest-negative Failed claim.
Getting started details
- The registry schema is centered on registry.csv with columns such as claim_id, entry_id, created, claim, status, location, dependencies, script, scope, supersedes, notes, active. The claim life cycle moves from Planned to Verified or Failed via the official runner. The system enforces that no folder exists without a registry row.
Releases and traction
- There are no releases listed beyond the latest 0 entry. The repository has 122 stars and 0 forks, with 0 open issues, and the last push on 2026-07-13.
What you can expect to use
- Commands to bootstrap, register, init folders, and run gates as shown in the Quickstart section.
- A compact, project-agnostic framework designed to be copied into existing repos, with a memory/indexing layer and an autonomous loop for progression governed by frozen gates.