Potpie is an open-source context graph platform for AI-native software development: it maps your codebase, docs, tickets and decisions into a living knowledge graph, then runs specialized AI agents — for debugging, code generation, QA and impact analysis — on top of that context. The project has earned more than 5,700 stars on GitHub and is actively developed, with its latest updates in 2026.
General coding agents are only as good as the context they can see, and in a large organization that context is scattered across repositories, Jira tickets, Confluence pages and people’s heads. Potpie connects those sources through approved APIs and builds a typed ontology of your engineering world — services, files, owners, policies and how they relate — refreshed on every commit. On top of the graph sits an agents layer with an orchestrator that plans, delegates and monitors execution: a debugging agent can trace why checkout latency regressed, an impact-analysis agent can report what breaks if a policy changes, and a QA agent can draft targeted regression tests, all with sources you can follow. Potpie is model-agnostic, working with Claude, GPT, Gemini or your own weights, and the open-source core can be self-hosted in your own VPC.
Potpie’s core is free and open source (Apache-2.0) to self-host. The company also offers a commercial enterprise platform and embedded forward-deployed engineering engagements for large organizations, priced by project.
Potpie is for engineering teams — especially at larger companies — whose AI coding tools underperform because they lack organizational context, and who want agents that understand their architecture, history and decisions.