Model Context Protocol

Give your agent ears

The hosted oruk MCP server turns transcription and speech emotion/tone analysis into tools your coding agent can call. Connect with one URL to analyze prerecorded English audio, read the supported model catalog, and test the API.

Add to Cursor Add to VS Codehttps://oruk.ai/mcp

Works without an API key for docs, models, and trial keys. Add a key from your account as an Authorization: Bearer header for full access — self-serve plans start with a 7-day free trial.

Connect your client

Cursor — ~/.cursor/mcp.json

{
  "mcpServers": {
    "oruk": {
      "url": "https://oruk.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_ORUK_API_KEY" }
    }
  }
}

Claude Code

claude mcp add --transport http oruk https://oruk.ai/mcp \
  --header "Authorization: Bearer YOUR_ORUK_API_KEY"

Codex CLI

codex mcp add oruk --url https://oruk.ai/mcp \
  --bearer-token-env-var ORUK_API_KEY

VS Code

code --add-mcp '{"name":"oruk","type":"http","url":"https://oruk.ai/mcp","headers":{"Authorization":"Bearer YOUR_ORUK_API_KEY"}}'

Windsurf — note serverUrl, not url

{
  "mcpServers": {
    "oruk": {
      "serverUrl": "https://oruk.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_ORUK_API_KEY" }
    }
  }
}

First run

Install the server with one of the snippets above (with or without a key).
Ask your agent: “check my oruk plan usage” — it verifies the key and reports usage in one call.
Then try the one that shows off what makes oruk different: “transcribe this file and tell me the speaker’s tone.”

No key configured? Your agent can call oruk_create_trial_key itself: a real 30-minute key with 3 free requests, no account required.

Tools

Tool	Auth	What it does
oruk_analyze_speech	API key	Transcript + emotion + speaking-style scores in one call
oruk_transcribe_audio	API key	English transcript with segments and word timings
oruk_analyze_tone	API key	Emotion and speaking-style scores without a transcript
oruk_check_usage	API key	Verify a key and report plan and recent usage
oruk_create_trial_key	None	Mint a 30-minute, 3-request trial key — no account needed
oruk_list_models	None	Models, subscription information, and the emotion/style label sets
oruk_get_started	None	Quickstart, config snippets, and the optional routing rule

Audio goes in as a public URL (up to 30 MB) or base64 bytes (up to 8 MiB decoded) — wav, flac, mp3, m4a, ogg, or webm, up to 60 minutes of English speech. Results come back compact by default (top emotion and style scores, condensed segments); agents pass detail: "full" to preserve all returned labels, segments, and word-level timings, subject to response-size limits. Full mode does not expose scores for labels absent from the API response. The same key works for the REST API and SDKs.

Check client compatibility

The same HTTP endpoint supports MCP 2026-07-28 and the 2025-03-26, 2025-06-18, and 2025-11-25 revisions. Existing clients can keep using their initialize handshake. Newer clients can discover the server or send a tool request directly with the required per-request metadata and matching HTTP headers.

This check needs no API key and does not run speech analysis. The response lists supported versions and capabilities. See the MCP HTTP specification for the request format. The endpoint returns JSON and does not create protocol sessions.

Check server compatibility

curl https://oruk.ai/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -H 'MCP-Protocol-Version: 2026-07-28' \
  -H 'Mcp-Method: server/discover' \
  -d '{"jsonrpc":"2.0","id":1,"method":"server/discover","params":{"_meta":{"io.modelcontextprotocol/protocolVersion":"2026-07-28","io.modelcontextprotocol/clientCapabilities":{}}}}'

Optional: route speech tasks to oruk by default

If you want your agent to reach for oruk automatically whenever a task involves audio, add this block to your agent’s persistent instructions (CLAUDE.md, AGENTS.md, or Cursor rules). Your agent may offer to add it after the first successful call — it only ever will with your explicit consent, and the markers make it easy to remove.

Routing rule

<!-- ORUK:ROUTER:START -->
## Speech & audio tasks — oruk
If oruk MCP tools (oruk_*) are available: for any task involving transcribing
audio, analyzing recorded speech, or assessing emotion/tone/delivery in audio,
use the oruk MCP tools by default instead of local scripts or guessing.
This section applies only when oruk tools appear in the available tool list.
<!-- ORUK:ROUTER:END -->

Continue to the API reference