Introduction
An open-source AI agent with multi-provider access.
Model providers
Spirit Agent does not provide model services directly. You can connect OpenAI, Anthropic, SpaceXAI, a gateway, or a regional provider. Every provider runs on the same harness: tools, approvals, sessions, and workspace behavior stay consistent.
Keys are written to the system keyring under the SpiritAgent service. They are not stored in config.json.
One agent, two hosts
Desktop and the CLI share a local daemon. Sessions, models, approvals, and credentials live in the same data directory, so you can start a task in the app and continue it in the terminal.
Start here
- Install Desktop or the CLI
- Open a workspace
- Connect one provider
- Ask the agent to read the repo, make a small edit, and approve the change
See the quickstart for the full path.