Your agent writes code fast.
paw makes it write code right.
An open-source Personal Agent Workforce: standards that are enforced, not suggested, 18 specialized agents, 7 skill bundles, git guardrails, and a CLI.
Install in under a minute
git clone https://github.com/Effulgent-Point/paw.git ~/paw
cd ~/paw
./paw setupTry the polished walkthrough here on the hosted tutorial, or run it locally after install with paw tutorial --web. Then run paw init inside a project to opt into the gates that make sense for the stack.
What is in the kit
An architect, a planner, a builder, reviewers, and more, each with a narrow job and the right permissions.
Reusable capability packs your agent loads on demand, from git doctrine to test-first workflows.
Gates that fire on real events, so standards are enforced by the tooling, not left to good intentions.
Shipped recently
paw is a moving target on purpose. Here is what landed on main in the last few arcs, and why each one earned its keep.
A new practical walkthrough for encoding your own doctrine, deployment targets, and persona agent. It is the page we wish someone had handed us before we started. Read it in order once, come back to the sections that match where you are stuck.
The catalog page enumerates the current kit end to end. Each agent card carries its phase and permission tier so you can see, at a glance, which ones can only look and which ones can write.
The browser walkthrough matches the CLI walkthrough one to one. Learn paw, agents, hooks, rules, the CLI, workflows, terminology, contributing, and the practice buggy app. Generated from a single source so the two never drift.
The /library page lays out the wider agent roster paw draws from, with in-paw tags on the ones that ship in the box and recipes to chain them. A menu to steal from, not a spec.
Build your own agentic harness.
paw is one implementation of a pattern, not the pattern itself. If you want to encode your own doctrine, your own hooks, and your own agents around your day-to-day, we wrote the playbook we wish we had. Deployment targets, corpus mining, when to encode a rule vs spawn an agent, all of it.