---
## What is this?
Agents are only as useful as what they can reach — and the services they need expose a
fragmented mess of interfaces: MCP servers, REST/OpenAPI, GraphQL, CLIs. That fragmentation
is fine. What's missing is the **map**: one place that says *"here is everything
`{service}` exposes, and here is exactly how to authenticate to each interface."*
**integrations.sh** is that map. Type a domain and get every publicly-reachable integration
surface for it, grouped by format, each annotated with:
- **where it lives** — endpoint, OpenAPI spec, transport, docs;
- **how you authenticate** — the credential you need, and the setup prose to acquire it;
- **how trustworthy the answer is** — every fact is tagged `detected` (re-verifiable from a
machine signal the service publishes) or `discovered` (read from the service's docs).
The page you read and the JSON your agent fetches are the same content.
## Why auth is the bet
Discovery takes a minute; **auth takes an hour**. Finding out that Stripe has an MCP server is
easy — figuring out which key it wants, where to mint one with least privilege, and which
header to put it in is the part that actually costs you. So every surface leads with a
grounded, cited credential guide rather than a link dump.
---
## Explore a domain
Every domain page groups a service's surfaces by format (**MCP · REST/OpenAPI · GraphQL ·
CLI**), followed by the shared **credentials** those surfaces reference — defined once, bound
per-surface.
