Mimir

Mimir provides an MCP server you run locally or in a container so your IDE can access structured playbooks, workflows, activities, and guidance. It enables you to query guidance, generate work plans, track progress, and evolve your playbooks as you work, all through your editor and a lightweight MCP client.

How to use

Interact with Mimir through your MCP client inside your IDE to access playbooks, workflows, activities, and howtos. You can ask for guidance on a specific scenario, request a work plan, or check progress against planned artifacts. Use the IDE’s MCP features to fetch activity guidance, create work orders, review improvements, and apply updates to playbooks as your project evolves.

How to install

Prerequisites: you need Docker installed to run the MCP server container. You also need Python 3.11+ if you run locally, and your IDE should support MCP (Windsurf, Claude Desktop, Cursor, or similar). Ensure you have an MCP user ready for login and access to the MCP workflow.

Step 1. Start the MCP server with persistent data using Docker.

# Pull the latest release from Azure Container Registry
docker pull acrmimir.azurecr.io/mimir:release-latest

# Run with persistent storage
docker run -d \
  --name mimir \
  -p 8000:8000 \
  -v ~/mimir-data:/app/data \
  -e MIMIR_USER=yourusername \
  -e MIMIR_EMAIL=you@example.com \
  acrmimir.azurecr.io/mimir:release-latest

# Access the web interface
open http://localhost:8000

Step 2. Configure MCP in your IDE. Choose the block that matches your IDE and paste the configuration snippet in the appropriate MCP config location. You will replace placeholder paths and user values with your actual setup.

Configuration for MCP clients

Use the following MCP configurations to connect your editor’s MCP client to Mimir. Each snippet sets up a local stdio server that runs inside a Docker container and exposes an MCP endpoint to your IDE. Replace yourusername with your actual username used in the container.

{
  "mcpServers": {
    "mimir": {
      "command": "docker",
      "args": [
        "exec",
        "-i",
        "-e",
        "MIMIR_MCP_MODE=1",
        "mimir",
        "python",
        "manage.py",
        "mcp_server",
        "--user=yourusername"
      ]
    }
  }
}

For Claude Desktop and Cursor, the same approach uses the following structures to connect the MCP server inside the container.

{
  "mcpServers": {
    "mimir": {
      "command": "docker",
      "args": [
        "exec",
        "-i",
        "-e",
        "MIMIR_MCP_MODE=1",
        "mimir",
        "python",
        "manage.py",
        "mcp_server",
        "--user=yourusername"
      ]
    }
  }
}

Note: You must replace yourusername with the actual username you use for Mimir. Your data persists in the mounted volume, so container updates will not erase your work.

Troubleshooting and notes

If the MCP server does not respond, run the server manually to test connectivity and ensure the user exists. Start the server with the same user you configured in the IDE and verify environment variables are set correctly. Restart your IDE after updating MCP settings to apply changes.

Security and maintenance

Keep your Docker image up to date, and rotate credentials periodically. Use a dedicated MCP user for your local editor connections and avoid exposing administrative accounts in shared environments.

Notes

Your MCP environment includes data stored in a persistent volume so you can upgrade the container without losing playbooks, workflows, or artifacts.

What you can do with the MCP tools

The MCP client can manage playbooks, workflows, and activities using a set of actions. You can create, list, get details, update, and delete playbooks and workflows, as well as create, list, get details, update, delete, and define dependencies for activities.

Available tools

create_playbook

Create a new draft playbook with metadata such as name, description, and initial version.

list_playbooks

List playbooks with filters for status like draft, released, or all.

get_playbook

Retrieve detailed information about a playbook, including its workflows and nested content.

update_playbook

Update playbook details; the version is auto-incremented.

delete_playbook

Delete a draft playbook to remove it from the library.

create_workflow

Add a new workflow to a chosen playbook.

list_workflows

List all workflows for a specific playbook.

get_workflow

Get details of a workflow, including its activities.

update_workflow

Update workflow details and metadata.

delete_workflow

Remove a workflow from a playbook.

create_activity

Add a new activity to a workflow with guidance and dependencies.

list_activities

List activities within a workflow.

get_activity

Get details for an activity, including dependencies.

update_activity

Update activity guidance, name, or phase.

delete_activity

Delete an activity from a workflow.

set_predecessor

Define dependencies between activities, validating against circular references.