# BizCrush API and MCP — Full Reference

## BizCrush MCP

### Connection

- MCP endpoint: https://extapi.bizcrush.ai/mcp
- Transport: Streamable HTTP, stateless
- Recommended authentication: OAuth 2.1 Authorization Code with S256 PKCE
- Protected resource metadata: https://extapi.bizcrush.ai/.well-known/oauth-protected-resource
- Authorization server metadata: https://extapi.bizcrush.ai/.well-known/oauth-authorization-server
- Human-readable guide: https://extapi.bizcrush.ai/developer/mcp

An OAuth-capable MCP client discovers authentication from the endpoint. Its browser is redirected to https://bizcru.sh, where the user signs in with their normal BizCrush account and reviews the requested permissions. No BizCrush API key is required or shown in this OAuth flow. Do not ask the user to create, paste, or share an API key for OAuth.

Static X-API-Key authentication remains available only for developer clients that do not support OAuth discovery and can store custom headers securely.

### OAuth scopes

- meetings:read — read meetings, transcripts, summaries, settings, and contacts
- meetings:write — create or change meetings and meeting data
- notion:write — upload meeting content to the user's connected Notion workspace

### Tools

The server exposes 22 tools dynamically, including meeting search and management, transcript and summary access, contact lookup, Slack summary delivery, and Notion upload. Discover the exact current schemas with the MCP tools/list method after authentication.

## Standalone APIs

> Standalone BizCrush APIs for speech, audio cleanup, and text translation.

## Text Translation

### POST /v1/translate-text
Auth: X-API-Key. JSON fields: text (required), target_languages (required list), source_language (optional, default auto).

## Audio Noise Reduction

### POST /v1/denoise
Auth: X-API-Key. JSON fields: audio_url (required), atten_limit_db and disable_adaptive_mix (optional). Returns denoised_audio_url, applied_atten_db, and frame_lsnr_db.

## Speech-to-Text API

### POST /v1/stt
Auth: ?api_key or ?authorization. JSON or protobuf. File transcription returns text, detected_language, confidence, duration_ms, and speaker utterances.

### WebSocket /v1/stt/stream
Auth: query parameter. Supports JSON or protobuf, pcm16 or Opus, diarization, language hints, optional noise reduction, interim/final chunks, and an empty-frame end marker.

## Text-to-Speech API

### POST /v1/tts
Auth: query parameter. Protobuf SessionTTSRequest/SessionTTSResponse. Returns a signed OGG/Opus audio URL.