MCP tool reference

Every tool below is callable by any MCP client connected to https://innernet.live/api/mcp. Authentication is by Authorization: Bearer <token>.

innernet_list_projects

List all your context maps. Returns slug, name, type, tagline, and last-updated date for each. Call this first to see what exists.

{}
innernet_load_project

Load a full context map by slug. Returns dimensions (markdown), nodes (JSON), recent commits, and the project's capture_protocol. Use at the start of a session.

{ "slug": "my-project" }
innernet_save_context

Create or update a project from structured content. Pass an object of named dimensions; each value is markdown. Re-call to update.

{ "name": "Brand reset", "type": "brand", "dimensions": { "vision": "...", "voice": "..." } }
innernet_capture

Quick capture mid-conversation. Lighter than save_context — appends a note to the project's capture queue without restructuring the full map.

{ "project_slug": "brand-reset", "content": "Just decided: ditch the all-caps wordmark.", "tags": ["design"] }
innernet_get_dimension

Read one dimension's full markdown. Use when you only need one section rather than loading the whole project.

{ "project_slug": "brand-reset", "dimension_name": "voice" }
innernet_get_capture_protocol

Read the per-project capture protocol — markdown instructions telling the AI what to save and how. Returned automatically inside innernet_load_project; call this when you want it refreshed mid-session.

{ "project_slug": "brand-reset" }
innernet_set_capture_protocol

Replace the project's capture protocol. Use when you and the user agree on a refined approach to capturing context for this project.

{ "project_slug": "brand-reset", "protocol": "# Capture protocol\n\n- Save voice rules verbatim..." }
search

Search across projects by keyword. Used by ChatGPT Deep Research.

{ "query": "brand" }
fetch

Fetch a project's full content as a single markdown blob. Used by ChatGPT Deep Research.

{ "id": "brand-reset" }

Calling pattern

All tools follow the JSON-RPC 2.0 MCP convention. Most users never call them directly — the AI client wraps them. If you want to script against the server, see Connect / custom clients.