# Open source tools from the Unblocked team URL: https://getunblocked.com/open-source/ We're building the tools we wish we had for context engineering. These are open source, battle-tested in our own product, and ready for your repos. --- - [Engineering Social Graph Builder](https://github.com/unblocked/engineering-social-graph) — Build a developer social graph from GitHub PR history. Builds a weighted directed graph of who reviews whose PRs, with temporal decay so recent activity matters most. Automatically clusters contributors into teams using community detection on mutual review patterns. Identifies domain experts for every code area using git history. Generates an interactive HTML visualization with force-directed graph, heat map, peer tables, team cards, and area expertise views. Flags single-owner code areas and knowledge silos with bus factor analysis. - [Document Query Engine](https://github.com/unblocked/document-query-engine) — Natural language → validated MongoDB aggregation pipelines over GitHub PR and issue data. A hands-on workshop project that goes beyond semantic RAG. Similarity search finds related documents, but it can’t reason about time, traverse relationships, or apply precise filters. This workshop teaches you to build a structured query engine that translates natural language into validated MongoDB aggregation pipelines — enabling agents to query the real data behind their workflows. Ships with a complete web UI (query playground + chat agent) where the engine starts as a stub you build to life across six guided steps: schema discovery → identity resolution → synthesis → validation → retry → full-text search. - [Context Engine Simulator](https://github.com/unblocked/context-engine-simulator) — A/B test whether pre-gathered context helps coding agents complete tasks faster, cheaper, and better. Runs the same coding task twice against a real repository — once as a baseline (the agent works from the task description alone) and once context-enhanced (the agent first gathers relevant context from the codebase and connected MCP sources, then plans and implements from that briefing). Both results are scored 0–100 by an independent evaluator agent against your acceptance criteria, and the tool reports the quality, speed, and cost difference. Works with multiple coding-agent CLIs — Claude Code, OpenAI Codex, Cursor, and Grok — so you can compare how each benefits from context. Results are output as a terminal table, a JSON file, and a standalone HTML report. - [Repo Rules Agent](https://github.com/unblocked/repo-rules-agent) — Extract and index AI coding instructions from rules files. Sweeps ~40 known rules-file conventions (CLAUDE.md, AGENTS.md, .cursorrules, .github/copilot-instructions.md, etc.), extracts structured records via LLM tool-calling, deduplicates similar rules across files using embeddings, flags contradictions for human resolution, and gives you a queryable index scoped to your task. Not “dump every rules file into context and hope” – instead: “give me the must-severity Python rules that apply to code review.” - [Claude Harness](/demo/) — A/B comparison harness for Claude Code: run tasks with and without Unblocked context, then compare results. Creates two isolated git worktrees from the same branch and runs Claude Code in parallel on both — one baseline (Unblocked tools blocked entirely) and one with Unblocked context (nudged to call context_research first and throughout). Captures diffs, token usage, cost, tool calls, and timing from both runs, then generates console, JSON, and standalone HTML comparison reports with head-to-head bar charts, per-arm detail cards, and tool-usage breakdowns. Built-in contamination guards ensure the baseline arm can’t accidentally reach Unblocked, and the context arm must actually use it. ## Get Started Stop feeding context. Start shipping code. - Start for free: - Book a demo: