Model Context Protocol

Đúng Chưa? MCP Documentation

Machine-readable and agent-native documentation for Đúng Chưa's Model Context Protocol (MCP) server. Discover product capabilities, acoustic DSP admission gates, methodology boundaries, and verified routes.

Server Overview

Đúng Chưa? provides a public, unauthenticated Model Context Protocol (MCP) server adhering to the 2025-03-26 protocol specification. It allows autonomous agents, IDE assistants, and LLM web crawlers to retrieve deterministic product facts, acoustic scoring limits, and dialect baselines without needing credentials or executing mutations.

HTTP Endpoint

POST https://dungchua.app/api/mcp

Streamable JSON-RPC 2.0 over HTTP with CORS enabled for all origins (*).

Well-Known Discovery

GET https://dungchua.app/.well-known/mcp.json

Standard static metadata manifest declaring capabilities and registered resources.

Authentication & Scope

Authentication: None

Public read-only. No learner data, voice recordings, scoring sessions, or billing mutations are exposed.

Supported Resources

The server implements resources/list and resources/read. The following 4 canonical resources are served in application/json:

dungchua://product application/json

Canonical product description, problem definition, value-first preview doctrine, and high-level positioning.

dungchua://methodology application/json

The 5-step practice loop, quiet-room recording boundaries, honest abstention doctrine (unclear recording ≠ pronunciation failure), and claim ceilings.

dungchua://specifications application/json

Exact acoustic DSP admission gates (minimum 10 dB SNR, 25 clean voiced frames, 0.45 voiced ratio, max 10 octave jumps), YIN F0 DTW scoring engine, and phonation veto rules.

dungchua://routes application/json

Full directory of canonical URLs including practice rooms (/practice/ma, /practice/hoi-vs-nga), 6-tone guides, methodology, and AI discovery endpoints.

JSON-RPC 2.0 Wire Examples

All POST requests to /api/mcp use standard JSON-RPC 2.0:

1. Initialize

Negotiates MCP protocol handshake and returns server info.

// Request:
curl -X POST https://dungchua.app/api/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "initialize",
    "params": {
      "protocolVersion": "2025-03-26",
      "capabilities": {},
      "clientInfo": { "name": "AgentResearcher", "version": "1.0.0" }
    }
  }'

// Response:
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "protocolVersion": "2025-03-26",
    "capabilities": {
      "resources": { "listChanged": false }
    },
    "serverInfo": {
      "name": "dungchua-product-discovery",
      "version": "1.0.0"
    },
    "instructions": "Read-only public product discovery. No learner data, recording access, scoring, tools, or mutations are exposed."
  }
}

2. List Resources (resources/list)

// Request:
curl -X POST https://dungchua.app/api/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "id": 2, "method": "resources/list"}'

// Response:
{
  "jsonrpc": "2.0",
  "id": 2,
  "result": {
    "resources": [
      {
        "uri": "dungchua://product",
        "name": "Đúng Chưa? product summary",
        "description": "Canonical, claim-safe description of Đúng Chưa?.",
        "mimeType": "application/json"
      },
      {
        "uri": "dungchua://methodology",
        "name": "Đúng Chưa? methodology and limits",
        "description": "Safe practice loop and product boundaries.",
        "mimeType": "application/json"
      },
      {
        "uri": "dungchua://routes",
        "name": "Đúng Chưa? canonical routes",
        "description": "Public product and discovery routes.",
        "mimeType": "application/json"
      },
      {
        "uri": "dungchua://specifications",
        "name": "Đúng Chưa? technical and operational specifications",
        "description": "Measurable DSP admission gates, acoustic metrics, dialect baseline, and commercial terms.",
        "mimeType": "application/json"
      }
    ]
  }
}

3. Read Resource (resources/read)

// Request:
curl -X POST https://dungchua.app/api/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 3,
    "method": "resources/read",
    "params": { "uri": "dungchua://specifications" }
  }'

// Response:
{
  "jsonrpc": "2.0",
  "id": 3,
  "result": {
    "contents": [
      {
        "uri": "dungchua://specifications",
        "mimeType": "application/json",
        "text": "{\n  "acoustic_admission_gates": {\n    "min_snr_db": 10,\n    "min_clean_voiced_frames": 25,\n    "min_clean_voiced_ratio": 0.45,\n    "max_octave_jumps": 10,\n    "max_hardware_sentinel_frames": 2,\n    "invalid_signal_verdict": "ABSTAIN"\n  },\n  "scoring_engine": {\n    "pitch_extraction": "YIN fundamental frequency (F0) tracking",\n    "alignment": "Dynamic Time Warping (DTW) in semitone-normalized space",\n    "phonation_analysis": "Acoustic glottal break/cutoff detection for Northern ngã [35ˀ] and nặng [21ˀ]",\n    "asr_role": "None (ASR transcription is not a tone judge)"\n  }\n}"
      }
    ]
  }
}

Supported Tools & JSON Schema (tools/list & tools/call)

In addition to read-only resources, the MCP server exposes 2 deterministic tools with full JSON Schema inputSchema definitions:

{
  "tools": [
    {
      "name": "dungchua_get_tone_guide",
      "description": "Retrieve acoustic pitch contour, phonation cue, and minimal-pair drill specifications for any of the 6 Northern Vietnamese tones (ngang, huyen, sac, hoi, nga, nang).",
      "inputSchema": {
        "type": "object",
        "properties": {
          "tone": {
            "type": "string",
            "enum": ["ngang", "huyen", "sac", "hoi", "nga", "nang"],
            "description": "Target Northern Vietnamese tone key."
          }
        },
        "required": ["tone"]
      }
    },
    {
      "name": "dungchua_evaluate_recording_gates",
      "description": "Evaluate whether a learner audio frame summary passes Đúng Chưa's acoustic DSP admission gates (SNR >= 10 dB, voiced frames >= 25, voiced ratio >= 0.45) or must ABSTAIN.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "snr_db": { "type": "number", "description": "Measured signal-to-noise ratio in dB." },
          "clean_voiced_frames": { "type": "integer", "description": "Count of clean voiced F0 frames." },
          "clean_voiced_ratio": { "type": "number", "description": "Ratio of voiced frames (0.0 to 1.0)." }
        },
        "required": ["snr_db", "clean_voiced_frames", "clean_voiced_ratio"]
      }
    }
  ]
}

Connecting from an MCP Client

To connect an MCP client (such as Claude Desktop, Cursor, or an Antigravity agent), add this entry to your client configuration:

{
  "mcpServers": {
    "dungchua": {
      "url": "https://dungchua.app/api/mcp"
    }
  }
}