Potpie

About Potpie

Potpie is an open-source AI harness and engineering context layer for large codebases. Rather than shipping one general coding assistant, it builds a structured representation of a repository and then exposes specialist agents on top of it — for codebase question-answering, debugging, root-cause analysis, test generation, implementation planning and pull-request review. The argument the product makes is that general models fail on large repositories not because they cannot write code but because they lack the organisation-specific context: the service boundaries, the historical decisions, the conventions that make one change safe and another change a regression. Potpie's context layer is designed to be that missing input, and it claims 63% on SWE-bench Lite along with deployments over codebases of 50 million-plus lines in regulated industries. The product surface is split across Specialists (the task-specific agents), Forge (building custom agents against your own context), Recipes (reusable automation templates) and Trace (visibility into what an agent did and why). It is open source with a public repository carrying more than 5,000 stars, which matters for teams who need to run the context layer inside their own environment rather than shipping proprietary source to a vendor. The target buyer is an engineering organisation automating SDLC workflows rather than an individual developer wanting autocomplete.

Key Features

Tags