A terminal chat client for Ollama that lets you interact with local models, persist sessions, and safely run tool-enabled commands.
- Install dependencies.
npm install
- Make sure Ollama is installed, running, and has at least one model available.
- Start Locopilot.
npm start
- Follow the prompts to confirm the Ollama host and port, pick a model, and open a session.
Locopilot stores its local state in config.json and locopilot.db in the current working directory, so sessions stay tied to the folder you launch it from.
If you want to skip command approval in a trusted environment, use YOLO mode:
npm start -- --yolo
npm start -- -y
YOLO=true npm startPowerShell:
$env:YOLO="true"; npm start| Feature | What you get |
|---|---|
| Local Ollama chat | Streams responses from a local Ollama instance and renders them as terminal-friendly markdown. |
| Persistent sessions | Saves chats and settings locally so you can resume, switch, or delete sessions later. |
| Safe tool execution | Supports run_command with approval by default, plus YOLO mode for trusted automation. |
| Developer-focused commands | Includes /settings, /model, /compact, /dump, /sessions, /delete, /nudge, /help, and /exit. |
| Web and file tools | Offers web_search, fetch_url, fetch_image, read_file, and write_file for agentic workflows. |
| Context management | Compacts long conversations and clamps runtime context to the selected model's reported limit. |
- Node.js v16+
- Ollama installed locally or reachable from the configured host/port
Local Ollama needs no authentication. For a remote or proxied Ollama
endpoint that requires a Bearer token, set apiKey on that provider in
config.json:
{
"providers": [
{
"id": "my-remote-ollama",
"name": "Remote Ollama",
"provider": "ollama",
"baseUrl": "https://ollama.example.com",
"apiKey": "your-secret-token",
"model": "llama3.2"
}
]
}The key is sent as an Authorization: Bearer <key> header on every
request to that provider. Providers without an apiKey field behave
exactly as before.
