Your AI tools, on open models. No rewrite.
Keep Claude Code and the OpenAI or Anthropic code you already have. Point it at Flynn and it runs on open-weight models, with Flynn choosing the model for each task.
# Flynn speaks Anthropic's native /v1/messages: point Claude Code straight at it.
export ANTHROPIC_BASE_URL=https://api.flynn.digitalfrontier.so
export ANTHROPIC_AUTH_TOKEN="$FLYNN_API_KEY"
# optional: route every request through the Flynn router
export ANTHROPIC_MODEL=flynn
claudeFlynn features
Keep the tools you use
Claude Code and the OpenAI and Anthropic SDKs work as they are. Change the address and the key, and they run on open-weight models.
Big jobs in one go
Send up to 10,000 requests as one batch, in OpenAI's or Anthropic's batch format, and collect the results within a 24-hour window. Anything that can't run is listed with the reason, never lost.
The right model for each task
Let Flynn choose. It learns which model handles each kind of task best for what it costs.
Fewer broken agent runs
Open models often fumble tool calls in small ways. Flynn fixes the slips it can prove are safe, so agents like Claude Code keep going.
Get early access to Flynn.
Tell us what you'd run on open models, and we'll get you into the preview.