Corbell
Corbell builds a living architecture graph of a multi-repository backend system from its code and past design documents, then uses that graph to generate specs that follow established patterns and to review proposed changes.
Corbell is aimed at teams whose features span five to ten repositories, where architectural decisions such as a messaging technology choice or an authentication pattern tend to be re-litigated because they live only in chat messages, wiki pages or the memory of engineers who have left. It builds a knowledge graph of the actual architecture from the code and the team's design history, then generates a new design specification that follows the patterns already established rather than inventing new ones.
Tasks pushed to Linear from a generated spec carry the exact method signatures, call paths and cross-service impacts an AI coding agent needs to work on them without further discovery. Corbell runs entirely from a local checkout against local repositories, with no server or cloud setup required, and supports Claude or GPT-4o as its underlying model along with export integrations for Notion, Linear and Jira.
It is written in Python and requires Python 3.11 or later.