Quickstart

Create an agent and start chatting in your terminal. You can write agics, flows, and caps later, as you need more control over how your agent works.

Install Toolang

Install Toolang with uv:

uv tool install toolang

You can also use pip install toolang in a Python environment. Toolang requires Python 3.11 or later.

Check the installation:

toolang --version
toolang --help

Configure models

Set an API key for a supported provider in your shell, or use a running local Ollama server. For example:

export OPENAI_API_KEY=...

Toolang discovers models from configured providers. Inspect the providers and available models:

toolang providers
toolang models

Use the provider list to check which credentials are missing. Choose an available model inside Chat with /model MODEL.

Create your agent

toolang new alice
toolang alice chat

The first command creates alice under ~/.toolang/agents/alice by default. The second opens terminal Chat; you do not need to start an agent server first. Type a message and press Enter to send it.

Inside Chat, use /help for commands and /keys for keyboard shortcuts. Use /models to see the allowed models and /model MODEL to choose one. Enter /exit to leave Chat.

To return to the latest thread:

toolang alice chat -t

See Chat in the terminal for threads, queued input, and session settings.

Try a shared agent

A shared .too file can be run directly from its URL:

toolang https://toolang.ai/dev.too chat

With uvx, you can try it without installing Toolang first. The same provider setup is needed:

uvx toolang https://toolang.ai/dev.too chat

To keep a local agent based on that definition:

toolang clone https://toolang.ai/dev.too alice

Choose another name if alice already exists.

Add capabilities

Add a skill to your agent and inspect its available caps:

toolang alice skill add briceyan/review
toolang alice caps

Continue with Create and run agents, or write your first agentic script.