THE RIGHT CONTEXT.FOR YOUR AI.
Find the repository files, history and architecture decisions relevant to a task, and give an AI assistant the context it needs.
See how it works ↓Local server in development
A recorded execution is available alongside the animation.
01 / IN MOTION
Pick up where you left off.
Inspect a recorded execution
ContextMCP / get_context
JSON ↗02 / The starting point
Retrieve code and decisions relevant to a task.
A change request rarely arrives with the right filename attached. Code, history and architecture decisions provide entry points to find.
03 / What it brings together
- 01Search repository files.
- 02Retrieve architecture decisions and Git history.
- 03Assemble size-limited context for a task.
04 / Under the interface
- The constraint
- Retrieve task-relevant excerpts without changing the repository or requiring an embedding engine.
- The decision
- The server runs locally in Rust and exposes read-only MCP tools. Lexical search can work without an embedding engine.
The recorded execution linked at the top of this page uses the real server on two portfolio source files and shows the excerpts returned for two specific requests.
05 / What this explores
Retrieving context without changing the repository.
Turn a request into usable information, connect a repository to an agent and keep control over the amount of context passed along.
Project status & presentation scope
Local server in development. The recorded execution uses two portfolio files and demonstrates excerpt retrieval without measuring improvements in agent performance.