Add an MCP server endpoint to expose structured project context for external tools

[tech489]

Problem or use case

The entireio/cli project captures AI agent sessions, checkpoints, and context alongside git commits to unify code with reasoning. While this approach stores rich context internally, there’s currently no standardized interface for external tools to query structured project and session context in a consistent format.

I propose introducing an MCP (Model Context Protocol) server component as part of the CLI that exposes project context in a structured API. External tools (AI assistants, IDE plugins, automation frameworks, code search utilities, etc.) would be able to fetch this context on demand without having to parse raw files or infer structure manually.

Desired behavior

Context API Endpoint:

Expose a local HTTP server, e.g., running on localhost:<port>, serving structured JSON.

Example endpoints:

GET /context/repo — basic project metadata and settings

GET /context/sessions — list of captured AI sessions

GET /context/checkpoints — checkpoint metadata

GET /context/files?session=<id> — files touched by a session

GET /context/prompts?session=<id> — prompts/responses timeline

MCP Compatibility:

Align the API with MCP standards so tools using the MCP ecosystem can “plug and play.”

Support versioning to evolve responses safely.

Security & Access:

Optionally include access controls, API tokens, or local socket tokens to prevent unintended access.

Proposed solution

No response

Alternatives or workarounds

No response