Quick Start

Quick Start

This page gets Recollectium installed, initialized, and doing useful work quickly.

1. Install

Linux and macOS:

curl -LsSf https://raw.githubusercontent.com/AlfonsoDehesa/recollectium/main/install.sh | sh

Windows PowerShell:

powershell -ExecutionPolicy Bypass -c "irm https://raw.githubusercontent.com/AlfonsoDehesa/recollectium/main/install.ps1 | iex"

The bootstrap installer installs uv if needed, installs Recollectium as an isolated tool, prepares the configured embedding model, refreshes stale embeddings when needed, and installs shell completion where safe.

2. Verify

recollectium --version
recollectium --help

If you installed with pip, pipx, or uv tool, run initialization once. Bootstrap installs run embedding maintenance automatically, but it is safe to run init again if you want to verify setup:

recollectium init

The first initialization downloads or prepares the configured local FastEmbed model cache. It also refreshes stale embeddings when existing memories need it. Time and size depend on the selected model, memory count, and network speed.

3. Add a user memory

recollectium add   --space user   --type preference   --content "I prefer concise technical answers."

4. Search user memory

recollectium search-user "concise answers"

Searches default to all user buckets. Add --type preference only when you deliberately want to narrow recall.

5. Add a workspace memory

recollectium add   --space workspace   --workspace-uid recollectium   --type decision   --content "Use SQLite for local memory persistence."

6. Search workspace memory

recollectium search-workspace   "local persistence"   --workspace-uid recollectium

7. Start the local API service

recollectium service start api
recollectium service status
recollectium service discover

The default endpoint is http://127.0.0.1:8765.

8. Stop the service

recollectium service stop

Where to go next

• Learn the model: Concepts
• Configure paths and service settings: Configuration
• Use every command: CLI Reference
• Run API or MCP services: Service Management
• Open the browser control plane: WebUI
• Connect an MCP client: MCP Server
• Call the local HTTP API: API Reference