Changelog
What we've shipped, newest first. No account needed.
Prompt caching stays on, on every plan. fusr/auto requests use each provider's prompt cache, a chat stays on the model whose cache is warm, and the dashboard shows what caching saved on input. An estimate, not a bill.
Pro and Team get an alert center with a bell in the header. Each alert has a fix you can apply in one click, like pausing or capping your FUSR.ai key. Get alerts by email, Slack, or Discord, with quiet hours.
Retired the chat-plan tools to focus on API keys.
Create a personal API token to read your usage, providers, and alerts from /api/v1. Add a webhook and FUSR.ai sends signed events to your URL.
On Pro and Team, fusr/auto learns which model goes well for you on each kind of task. A thumbs up or down in the Routing log teaches it faster. Your rules still win, and it never picks a model over 3 times the price of FUSR.ai's pick.
Make a FUSR.ai key and point any OpenAI-compatible tool at fusr/auto. It picks a connected model by rules you can read and change on the Routing page. If that model errors, the next one takes the request.
Create a team and invite people by email as owners, admins, or members. Share company keys, give everyone their own FUSR.ai key, set a monthly budget and per-person caps, and see spend by person and by model.
After signup, three steps: connect a provider, open the dashboard, and send a first request. Dismiss it when you want.
The security page lists what the database keeps. Prompts and payment details are not stored. A pasted provider key is stored encrypted only in the LiteLLM proxy.