Self-host the community edition.

Free to run yourself, offered since August 2025 (formerly MiniMe MCP). Install the MCP server with Docker, run the setup wizard, and connect Claude, Cursor, VS Code, or Windsurf in about 5 minutes. Want it managed instead? Recallium Cloud is rolling out early access.

Quick Start

Three commands to a running engine

1

Install Docker

Recallium runs in a container - Docker is the only prerequisite.

Get Docker
2

Clone the repo

Pull the repository to your machine.

recallium-ai/recallium
3

Run the start script

./start-recallium.sh on macOS / Linux

start-recallium.bat on Windows

Your browser opens automatically at http://localhost:9001 - continue in the Setup Wizard below.
NOTEOn Windows with Ollama? See firewall setup in Troubleshooting first.
Setup Wizard · localhost:9001

Pick a model provider and you're done

The wizard walks you through three short steps in the browser. Bring your own API key, run fully local with Ollama, or connect an enterprise provider.

The Setup Wizard opens in your browser at localhost:9001
The Setup Wizard opens in your browser at localhost:9001
1

Choose your LLM provider

Test the connection before adding a provider. Recallium supports cloud, local, and enterprise backends.

Cloud Providers
Anthropic RECOMMENDED
Claude 3.5 Sonnet, Opus, Sonnet, Haiku
OpenAI
GPT-4o, GPT-4 Turbo, GPT-3.5 - function calling, streaming, vision
Google Gemini
Gemini 1.5 Pro & Flash - multi-modal
OpenRouter
100+ models, unified billing
Local · Offline
Ollama FREE
Llama 3, Mistral, Qwen, or any local model

Runs entirely on your machine - no API costs, fully private.

Enterprise
Palantir AIP
Models via the Foundry platform

Enterprise-grade and compliance-ready for corporate environments.

Test the connection before adding a provider
Test the connection before adding a provider
2

Set provider priority

Order your providers so Recallium fails over automatically if one is unavailable.

3

Done

Recallium's MCP tools are now available to every connected IDE and chat app.

Configure provider priority for automatic failover
Configure provider priority for automatic failover
IDE Integration

Connect your IDE & chat apps

Point your client at the local MCP server. Here's Cursor - the same block works across editors.

Cursor

Open Settings → Cursor Settings → MCP → Add new global MCP server, then paste:

mcp.json
{
  "mcpServers": {
    "recallium": {
      "url": "http://localhost:8001/mcp",
      "transport": "http"
    }
  }
}
Claude CodeCLI

Add the server, then restart your session:

terminal
claude mcp add --transport http recallium http://localhost:8001/mcp
VS CodeEditor

Press Ctrl/Cmd + Shift + PMCP: Add ServerHTTP → enter http://localhost:8001/mcp. Or add to your MCP config file:

settings.json
{
  "mcp": {
    "servers": {
      "recallium": {
        "type": "http",
        "url": "http://localhost:8001/mcp"
      }
    }
  }
}
Claude DesktopChat App
  1. Find the .jsonb extension file in claude-desktop-extension/ of the cloned repo
  2. Open Claude Desktop → SettingsExtensions
  3. Click Browse ExtensionsAdd Custom
  4. Select the .jsonb file to install
  5. Go to Connectors → Recallium and enable the tools you want
Migrate from MiniMe

Bring your memories across

The community edition has been offered since August 2025, first under the name MiniMe MCP. Moving a MiniMe install into Recallium takes two steps - your memories, projects, rules, and tasks all come with you.

Export from MiniMe
  1. Open the MiniMe UI
  2. Go to Projects → Export
  3. Export all memories - downloads as a .zip
Import into Recallium
  1. Install Recallium via the Quick Start
  2. Open the UI at http://localhost:9001
  3. Go to Projects → Import
  4. Upload the MiniMe .zip

Your memories, projects, rules, and tasks will all be imported. You can continue right where you left off.

Operations

Upgrades and backups

Upgrading is the same script you started with, and it keeps your data. Export first anyway, so a bad upgrade costs you nothing.

Upgrade
terminal
# macOS / Linux
./start-recallium.sh

# Windows
start-recallium.bat

The script pulls the latest version and restarts with your existing data intact. See what changed in the changelog.

Back up first
terminal
# macOS / Linux
./download-memories.sh

# Windows
download-memories.bat

Writes a dated ZIP holding every memory, project, rule and task, plus the computed embeddings so an import needs no re-processing. The same export lives in the web interface under Data Management, where you can scope it to one project or a date range.

Moving to a new machine

On the old machine
terminal
./download-memories.sh
# then copy the ZIP across
On the new one
  1. Install Recallium there first
  2. Open Data Management → Import
  3. Select the ZIP and choose Merge
  4. Click Import
Import modes
Merge
Adds to what is already stored. The one to use.
Replace
Clears everything first, then imports.
Fresh
Only runs against an empty database.
Project mapping

Each incoming project can either create a new project of the same name, or merge into one that already exists. That is how you consolidate projects, or rename them, as part of the move.

Export from the API

terminal
# everything
curl -o "backup.zip" http://localhost:8001/api/data/export

# one project
curl -o "project.zip" "http://localhost:8001/api/data/export?project_id=5"

# preview without downloading
curl http://localhost:8001/api/data/export/preview
Troubleshooting

Common fixes

Windows + Ollama setup

On Windows, Docker containers need your machine's IP address to reach Ollama.

Step 1: Configure the Ollama environment variable

  1. Search Start menu for "Edit the system environment variables"
  2. Click "Environment Variables"
  3. Under "System variables" → click "New"
  4. Name: OLLAMA_HOST → Value: 0.0.0.0:11434
  5. Click OK → restart Ollama

Step 2: Add a Windows firewall rule (run as Administrator):

powershell
netsh advfirewall firewall add rule name="Ollama" dir=in action=allow program="C:\Users\<YOUR_USERNAME>\AppData\Local\Programs\Ollama\ollama.exe" enable=yes profile=private

The start-recallium.bat script will auto-detect your IP address.

Port conflicts

Edit install/recallium.env before starting:

recallium.env
HOST_UI_PORT=9001        # Web UI
HOST_API_PORT=8001       # MCP API
HOST_POSTGRES_PORT=5433  # PostgreSQL

Then restart: ./start-recallium.sh

Docker Compose (advanced)

Prefer to manage it yourself? From the repo:

terminal
cd install
docker compose --env-file recallium.env pull
docker compose --env-file recallium.env up -d
Tools not appearing in IDE
  1. Verify the container is running: docker ps -f name=recallium
  2. Check health: curl http://localhost:8001/health
  3. Fully restart your IDE after config changes
Ollama connection issues
terminal
# Ensure Ollama is running
ollama serve

# Verify models are downloaded
ollama list

# Test connection
curl http://localhost:11434/api/version
Reset everything data loss

⚠️ This permanently deletes all data:

terminal
cd install
./reset-recallium.sh   # Linux/macOS
# or
reset-recallium.bat    # Windows
Management commands
terminal
# View logs
docker logs -f recallium

# Stop
docker stop recallium

# Restart
docker restart recallium

# Update to latest version
cd install && ./start-recallium.sh

Everything your agents need to remember.

Star the repo, then dive into the help guide and benchmarks.