FAQ

Installation

Does asyncat need internet to install?

The installer clones the repo from GitHub and runs npm install, so you need internet once for setup. After that, asyncat runs entirely offline if you use a local model provider.

What are the system requirements?

Node.js 20.19+, 22.13+, or 24+ and git are the only hard requirements. Python 3.10+, ffmpeg, and a C++ compiler are optional — needed only if you want local AI via llama.cpp or audio transcription.

How do I update asyncat?

Pull the latest changes from git, update dependencies, and rebuild native modules in the repository:

git pull
npm install
npm run electron:rebuild

How do I uninstall?

Simply delete the cloned repository directory. If you want to clean up all settings and database files, delete the data directory located at ~/.asyncat.

Models & providers

Can I use asyncat without a cloud API key?

Yes. asyncat works fully offline with local models via llama.cpp, Ollama, or MLX. No API key, no cloud, no data leaving your machine.

Which models are supported?

Any model accessible via an OpenAI-compatible API endpoint. Cloud providers include OpenAI, Anthropic, Gemini, Groq, DeepSeek, Together, Perplexity, Mistral, Cohere, Fireworks, Cerebras, OpenRouter, and more. Local engines: llama.cpp, Ollama, MLX, LM Studio.

Does asyncat support extended thinking / reasoning models?

Yes. asyncat extracts and displays reasoning tokens from Claude and o-series models. Reasoning effort is configurable (auto / low / medium / high) per session.

Can I switch models mid-session?

Yes — change the provider or model in Settings at any time. The new model applies from the next message onwards.

Data & privacy

Where is my data stored?

Everything is stored in SQLite at data/asyncat.db inside the install directory. Notes, tasks, calendar events, conversation history, agent memory, scheduled jobs — one file, on your machine.

Does asyncat send telemetry?

No. asyncat sends nothing anywhere unless you configure a cloud AI provider, in which case your prompts go to that provider according to their privacy policy.

Is asyncat multi-user?

Solo mode (default) uses a single local admin account. Set SOLO_MODE=false in den/.env for team server mode with user management and role-based access.

How do I back up my data?

Copy or backup the data/asyncat.db file located inside your application workspace directory.

Extending asyncat

Can I add my own tools?

MCP (Model Context Protocol) servers are the primary way to add tools. Register any MCP server in the web UI under Settings → MCP Servers (or by editing data/mcp.json) and its tools become available in every agent session.

Can I add my own skills?

Yes — create and edit skills in the Skills browser in the web UI, or place Markdown files directly in the ~/.asyncat/skills/ directory on your filesystem. See Skills.

Can I create custom agent profiles?

Yes. Profiles let you create named agent personas with their own system prompt, soul, skill set, and tool restrictions. Useful for specialized roles like "code reviewer" or "DevOps bot". See Profiles.

Can I access asyncat programmatically?

The den backend exposes a REST + WebSocket API on port 8716. Any agent run can be triggered via POST /api/ai/run. See API docs.