AI product Open source
Unblocked is a context engine for agentic software development. It gives coding agents permissions-aware context from source code, pull requests, conversations, documentation, issues, and production or incident history, helping them use the reasons behind code changes rather than code search alone.
Its open-source document query engine uses automatic schema discovery, identity resolution across systems, plain-language-to-database query generation, validation that blocks dangerous operators, error-fed retries, and full-text search. The project also includes a context-engine simulator for testing whether pre-gathered context improves agent tasks. In the demonstrations, attaching Unblocked context to Claude sessions reduced token cost by about half and helped trace a regression to a model switch before creating a fix pull request.
3 uses taken from transcripts — each links to the moment in the video.
A context engine that pulls together scattered institutional knowledge — source code, PRs, Slack conversations, docs, incident history — and delivers grounded, task-specific, permissions-aware context to coding agents. In demos, attaching it to Claude sessions cut token cost roughly 50% and helped trace a regression to a model switch, then create a fix PR.
Builds a context engine that delivers organizational context to humans and agents by ingesting and connecting scattered engineering context — PRs, review comments, tickets, wikis, docs, and Slack conversations — so it can be queried both semantically and structurally (by author, date, state, and other relationships). It continuously ingests content via webhook events and distills PR comments into best-practice memories.
Context management layer mentioned for storing rules and skills outside the repo
3 in the library.