CodePawl makes coding agents work together.
We build coordination infrastructure for coding agents: reviewable plans, guarded changes, traceable evidence, memory, replay, and cloud workflows.
Our first public product is Openpawl, an open runtime for coding-agent coordination. It starts in GitHub Actions and turns agent work into plans, validations, evidence bundles, and safety-gated changes.
- Openpawl — open runtime for coordinated coding agents
- TracePawl — failure diagnosis, replay, and postmortems for agent runs
- Mempawl — persistent operational memory for agentic systems
- CachePawl — optimization for long-horizon agent workloads
The current priority is Openpawl: a public GitHub Action and local runtime for reviewable coding-agent workflows.
Openpawl focuses on questions such as:
- What did the agent plan?
- Which files and context did it use?
- What validation ran?
- What changed, if anything?
- Can humans and other agents trust the run evidence?
CodePawl is interested in four core problems:
- How coding agents coordinate safely
- Why long-running agent work fails
- How agents preserve useful operational memory
- How agent workloads become cheaper, faster, and more reliable
CodePawl is early-stage research and engineering work. APIs, architecture, and product boundaries may change quickly.
Founder: An Nguyen Email: founder@codepawl.com Website: https://codepawl.com