Claudama

Use your Claude subscription as a Raycast AI provider — instead of paying per-token.

claudama wraps the claude CLI in an Ollama-compatible HTTP server, so any tool that speaks Ollama (Raycast first and foremost) can talk to Claude through your existing Claude Code subscription.

Quick start

1. Install via Homebrew:

   brew install atelpis/tap/claudama
   brew services start claudama   # one-time; auto-starts on login from here on

2. Point Raycast at claudama: in Raycast Settings → AI → Ollama set the host to http://127.0.0.1:11435, then click Sync Models. You should now see Claudama Haiku, Sonnet, and Opus in the list of models.

   

3. Wire up Raycast AI: assign them to your Quick AI, Chat, and Commands as you like.

   

That's it — Raycast AI now runs on your Claude subscription.

Alternative installation

From source:

go install github.com/atelpis/claudama/cmd/claudama@latest
claudama

Requires an installed and authenticated claude CLI on $PATH.

Configuration

Config is file-only. Lookup order — first existing file wins:

1. ~/.config/claudama/conf.toml
2. /opt/homebrew/etc/claudama/conf.toml
3. /usr/local/etc/claudama/conf.toml

Only the fields you specify override the defaults.

Full default config:

# Port claudama listens on (127.0.0.1 only).
port = 11435

# Absolute path to the `claude` binary. Leave empty to resolve via $PATH.
# Set this when running under `brew services`, where launchd does not inherit
# a useful PATH.
claude_path = ""

Notes

• Raycast only exposes its Ollama integration when the Ollama app is installed. You don't have to run it — the app just has to be present for Raycast to surface the provider slot.
• Raycast's Ollama slot points at a single host. claudama defaults to port 11435 so it coexists with Ollama on 11434; switch the host in Raycast settings to flip between them.