# infrai

> infrai gives your app 18 production backend modules — AI, video,
> email, SMS, scheduling, storage, payments and more — behind one API key, one
> wallet and one bill. It is a ZERO-INSTALL REST API: there is no SDK. Call it
> over HTTP from any language. Pricing is transparent and per-model — the final
> price you pay (plus any discount vs the vendor's public price) is live at
> GET /v1/pricing/models. Every response carries cost/vendor/latency metadata;
> every error carries a machine-readable hint and a retryable flag.

## How to call it (no SDK)

Every capability is a POST (or GET) to `https://api.infrai.cc/v1/...` with a
Bearer key and JSON body. AI inference is OpenAI-compatible — call the standard
`/v1/chat/completions` endpoint with `model: "auto"` for smart routing:

```bash
curl -X POST https://api.infrai.cc/v1/chat/completions \
  -H "Authorization: Bearer $INFRAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "auto", "messages": [{"role": "user", "content": "summarize this PR"}]}'
```

Responses use the envelope `{ ok, data, error, metadata }`. To get a key, sign
in at the console with Google/GitHub for a project key + $2 free credit (email
sign-in starts at $0). On `402 INSUFFICIENT_CREDIT`, POST
`https://api.infrai.cc/v1/account/topup` and open the returned `checkout_url`.

## 18 GA modules (298 REST routes)

- `ai-runtime` (https://api.infrai.cc/v1/ai) — ai.chat, ai.embed, ai.image
- `ai-video` (https://api.infrai.cc/v1/video) — video.generate, video.status, video.cancel
- `comm-email` (https://api.infrai.cc/v1/email) — email.send, email.list, email.get
- `sms` (https://api.infrai.cc/v1/sms) — sms.send, sms.otp, sms.verify
- `scheduling` (https://api.infrai.cc/v1/queue) — cron.create, cron.list, queue.publish
- `observability` (https://api.infrai.cc/v1/errors) — errors.capture, metrics.report, flags.is_enabled
- `captcha` (https://api.infrai.cc/v1/captcha) — captcha.verify
- `pdf` (https://api.infrai.cc/v1/pdf) — pdf.generate, pdf.merge, pdf.split
- `image-process` (https://api.infrai.cc/v1/image) — image.process, image.metadata
- `realtime` (https://api.infrai.cc/v1/realtime) — realtime.token.issue, realtime.channel.create, realtime.publish
- `storage` (https://api.infrai.cc/v1/storage) — storage.bucket.create, storage.bucket.list, storage.object.presign
- `analytics` (https://api.infrai.cc/v1/analytics) — analytics.track, analytics.identify, analytics.query.events
- `auth` (https://api.infrai.cc/v1/auth) — auth.user.create, auth.session.create, auth.session.verify
- `rtc` (https://api.infrai.cc/v1/rtc) — rtc.room.create, rtc.token.issue, rtc.room.list
- `db` (https://api.infrai.cc/v1/db) — db.project.create, db.branch.create, db.snapshot.create
- `dns` (https://api.infrai.cc/v1/dns) — dns.domain.add, dns.domain.verify, dns.record.create
- `risk` (https://api.infrai.cc/v1/risk) — risk.score, risk.device.fingerprint, risk.event.report
- `billing` (https://api.infrai.cc/v1/account) — account.balance, account.usage, account.topup

## API hosts

- Western: https://api.infrai.cc
- China: https://cn.infrai.cc

## For AI agents & tools (machine discovery)

If you are an AI coding assistant or agent, call infrai directly over HTTP, or
use the MCP server:

- MCP server (one command, all 18 modules as tools): `npx @infrai/mcp-server`
  Works with Claude Desktop, Claude Code, Cursor, Cline, Continue, Windsurf, Aider.
- OpenAPI 3.1 spec (generated from SSOT, with request schemas + examples):
  https://infrai.cc/.well-known/openapi.json
- Discovery manifest (full capability table: method/path/params/examples/errors):
  https://infrai.cc/.well-known/infrai-discovery.json
- MCP registry descriptor: https://infrai.cc/.well-known/mcp.json

The API is self-describing — never guess a parameter name or an allowed value.
Use the live discovery endpoints to read the exact schema before building a call:
- `GET https://api.infrai.cc/v1/discovery` — capability index (id/method/path/availability/billing). Lightweight, no params.
- `GET https://api.infrai.cc/v1/discovery/{capability}` — per-capability detail. `params` is the full request JSON Schema: `params.properties.<field>.enum` is the exact allowed-value set for any enum field (e.g. `account.webhooks.register` → `events[]`, `sms.send` → `route_class`/`mode`), and nested array elements carry their own `items` schema. `dynamic_params` lists runtime choices (e.g. the `vendor` enum reflects the providers your account has keys for). `response` is the result shape for chaining.

Every documented capability ships runnable curl / Python (requests) /
JavaScript (fetch) examples in the discovery manifest and the docs.

## Links

- Site: https://infrai.cc
- Docs: https://docs.infrai.cc