One API for models and tools
Setu puts AI models, the apps your users already use, and the controls that keep actions safe behind one key. Every call is checked and logged.
product-overviewHow it fits together
- 1
Models
Your app or agent thinks with a model behind the OpenAI-compatible API, or speaks and listens with the voice API.
- 2
Tools
It finds the right tool, your user connects the app once, and Setu runs the action for that user.
- 3
Controls
Every call passes the same checks, writes wait for approval, and you hear about what happened through webhooks.
All building blocks
Each part shows its real status.
Models API
BetaOpenAI-compatible chat completions with streaming and tool calls. Point your OpenAI client at Setu and keep your code.
Learn more about Models APIToolkits and actions
BetaRun actions in Gmail, Calendar, Drive, Docs, Tasks, Sheets and more. Write actions wait for approval.
Learn more about Toolkits and actionsTool search
BetaDescribe the task in plain words and get the right tools back, so your agent does not carry the whole catalog.
Learn more about Tool searchConnect links
BetaSend your users a link to connect their own apps. Setu keeps the tokens and tells you when a connection breaks.
Learn more about Connect linksMCP
BetaCall Setu tools from any MCP client, and bring outside MCP servers in as toolkits.
Learn more about MCPWebhooks and triggers
BetaGet app events, finished runs, approvals and spend alerts sent to your server, with replay for failed deliveries.
Learn more about Webhooks and triggersVoice
BetaText to speech with Nepali and Hindi, and speech to text, on the same key as your models.
Learn more about VoiceSafety and controls
BetaScoped keys, approvals, spend caps, rate limits, an audit log and data retention on every call.
Learn more about Safety and controls