One engine, five framings.
Each project type wires a different surface and a different sidebar but shares the same provider routing, guardrails, audit, and cost dashboard.
Pick the right type at create-time
The project_type is fixed for the life of the project. The Setup wizard makes the right choice obvious.
Fully-controlled AI endpoints — fixed prompt, schema validation, sessions, streaming, failover.
Egress gateway for Claude Code / Codex / Cursor / Aider. PII redaction + secret scanner default-secure.
HTTP proxy with method/header policies, OAuth injection, SSRF guard, rate limits.
Aggregate multiple upstream MCP servers behind one endpoint with health checks + auth.
Expose any AI Gateway endpoint as an MCP tool — agents call your endpoints via JSON-RPC.
Not sure which to pick?
Every detail page has a 'use this when / pick something else when' block. Or jump straight to the docs.