Integrations
Integrations
Most tools that already speak to a model vendor can speak to BeatAPI instead. Nothing is rewritten — you change a base URL, paste a BeatAPI key, and pick a model ID.
BeatAPI answers on four request formats, so a tool keeps using whichever one it was built for.
AI coding tools
Terminal coding agent, configured through ANTHROPIC_BASE_URL.
Add BeatAPI as a custom model provider in config.toml.
Two environment variables point it at the native Gemini endpoint.
The VS Code extension, using its OpenAI Compatible provider.
Override the OpenAI base URL in the editor’s model settings.
Manage Claude Code and Codex CLI providers from one desktop app.
Chat clients
Add BeatAPI as an OpenAI-type provider and load the model list.
Works through the OpenAI, Claude, or Gemini provider types.
Use BeatAPI as the chat model behind a local knowledge base.
Platforms
Configure BeatAPI as a model provider for apps, agents, and workflows.
Add BeatAPI as a custom translation service in the browser extension.
What every guide has in common
Three values are the same everywhere, so if a tool is not listed above, try its OpenAI-compatible option with these:
The base URL is the only place the /v1 suffix is ambiguous. Tools that follow the OpenAI SDK convention want https://api.beatapi.io/v1; tools built on the Anthropic SDK want https://api.beatapi.io and add /v1/messages themselves. Each guide states which one applies.
Choosing a model
Every integration reaches the same catalogue with the same key. These are reasonable starting points; the Text API pages list each family in full.
Keeping keys safe
- Create a separate key per tool so one can be revoked without disturbing the others.
- Prefer an environment variable over a value pasted into a config file that may be committed.
- Add config files that hold a key to
.gitignore. - Watch spend per key under
by_api_keyinGET /v1/usageor on the dashboard.