StudyLoopGitHub

Agent integration

Let your AI agent use StudyLoop’s local study context.

Model Context Protocol (MCP) gives a supported agent structured tools for reading study context and recording real outcomes. The server runs locally over standard input/output.

What it enables

Structured study actions instead of guessed shell commands

Understand the starting point

List courses, inspect due work, see available session targets, and retrieve study history.

Create review material

Turn a local chapter into flashcards or quiz questions and save them in StudyLoop’s expected format.

Record progress

Record a card result or update a backlog topic after the learner has actually practised.

Manage a live session

Check the available session choices and end the active session through the same local authority as the Web UI.

Connect it

Use the installed command, or point at a source checkout

If the StudyLoop installer put studyloop-mcp on your path, this is the smallest generic MCP configuration:

MCP configuration
{
  "mcpServers": {
    "studyloop-mcp": {
      "command": "studyloop-mcp",
      "args": []
    }
  }
}

For a development checkout, keep the path absolute and replace the example:

MCP configuration from source
{
  "mcpServers": {
    "studyloop-mcp": {
      "command": "uv",
      "args": [
        "run",
        "--project",
        "/absolute/path/to/studyloop/packages/studyloop",
        "studyloop-mcp"
      ]
    }
  }
}

The server uses your StudyLoop configuration

It opens the same local database and study paths as the CLI and Web UI. Test with non-sensitive material first.

Representative workflow

A useful agent turn has a visible beginning and end

  1. 1
    list_session_options

    Find what is actually available

    The agent sees real courses, lessons, topics, session types, and installed agents instead of inventing choices.

  2. 2
    get_study_context

    Check due work and recent progress

    This gives the conversation a starting point without treating notes as proof of learning.

  3. 3
    get_study_history

    Reconnect with earlier attempts

    The agent can notice wins and struggles on the chosen topic before asking the next question.

  4. 4
    record_topic_progress

    Write only after evidence exists

    Record an outcome when the learner has explained, retrieved, built, or debugged something.

Current boundary

MCP does not manage Study Plans yet

Create and edit plans through the Web UI or CLI. Keeping that boundary visible prevents an agent integration from appearing more connected than the shipped product really is.

Source of truth

Inspect the server and tests

The public repository contains the registered tools and a real stdio protocol smoke test.

Browse MCP source →