OpenLore

OpenLore helps developers keep AI coding agents grounded in large or unfamiliar repositories. It maps code relationships and retains architectural decisions so agents can find relevant context and check proposed changes.

• Task context: Finds functions, callers, tests and specifications relevant to a task, with application code and infrastructure definitions in the same graph.
• Persistent memory: Keeps notes and decisions attached to code across renames and moves, updates the index as files change, and flags stale information.
• Change checks: Identifies potentially breaking exports and affected consumers, detects new paths into declared sensitive boundaries, and can block commits on selected findings.
• Code investigation: Traces execution paths, finds unreachable code, selects tests affected by a change and checks claims against code with citations.

OpenLore is open source under the MIT license and runs on macOS, Windows and Linux. Its analysis and queries work offline without an account or API key; source code and the index stay on your machine. It connects through MCP to coding assistants including Claude Code, Cursor, Cline and Continue. Optional semantic search uses a local CPU-based embedding model.

OpenLore analyzes and checks code but doesn't edit it. Its static analysis can miss reflective calls and plugin relationships without visible bindings. Small, familiar repositories with simple queries may incur more overhead than using the agent's built-in search.