Reference
FAQ.
Short answers to the questions developers usually ask before building their first Orchet agent.
Do I send Orchet my database schema?
No. Orchet stores contracts, not your DB schema. Developers own their databases; Orchet stores public contract snapshots, approval state, auth metadata, scopes, runtime listing metadata, and the tool schemas needed to plan safe calls.
Where do agents run?
In your infrastructure. Vercel, Fly, Render, AWS, GCP, and any HTTPS host are fine. Orchet calls your approved endpoints.
Where do provider secrets live?
Provider API keys stay in your agent. OAuth client credentials for Orchet-initiated connect flows are configured in Orchet's environment during approval and are not submitted as payload fields.
Do users browse the Orchet Store to connect apps?
They can browse for trust and discovery, but task flow is chat-first. If the user asks for a missing capability, Orchet should show the install or connect card directly in chat.
Which protocol should I choose?
Choose SDK/API for new agents and strict schemas, remote MCP if you already have an MCP server, and A2A when your service already exposes an autonomous peer with an Agent Card.