"Tool Discovery for AI agents - with the ability to call what you discover"

DCAP is a decentralized protocol enabling autonomous agents to discover, evaluate, and acquire computational capabilities at runtime through semantic broadcasting. Version 2.5 introduces structured connector information, allowing agents to not just discover tools, but autonomously invoke them.

Latest Specification: DCAP.md - Version 2.5 (October 2025)

For automation/parsing: Always fetch DCAP.md - it's a stable URL that always points to the latest version.

✅ Dynamic tool acquisition with structured connector object
✅ MCP transport support (stdio, SSE, HTTP streaming)
✅ Structured authentication (x402, bearer, API key, none)
✅ Protocol specification (MCP, REST, gRPC)
✅ Autonomous capability invocation without manual configuration

Added connector object to semantic_discover:

{
  "connector": {
    "transport": "http",
    "endpoint": "https://robonet.example.com:8080/mcp",
    "auth": {
      "type": "x402",
      "required": true,
      "details": {
        "network": "base-sepolia",
        "asset": "0x036CbD...",
        "currency": "USDC"
      }
    },
    "protocol": {
      "type": "mcp",
      "version": "2024-11-05"
    }
  }
}

Key Points:

• connector.transport: MCP transport type (stdio/sse/http)
• connector.endpoint: WHERE to connect (URL or command)
• connector.auth: Authentication requirements with structured details
• connector.protocol: Protocol type, version, and methods
• The unlock: Agents can now discover AND invoke tools without pre-configuration

Full changelog: See archive/v2.5.md

All historical versions are archived for reference:

• v2.5 (October 2025) - Dynamic tool acquisition with connector object
• v2.4 (October 2025) - Format-agnostic agent identification
• v2.3 (October 2025) - Payment attribution with ctx.payer
• v2.2 (October 2025) - Cost tracking and economic efficiency
• v2.1 (October 2025) - Chain pattern detection with ctx.args
• v2.0 (September 2025) - Initial public specification

Recommended: Implement DCAP.md (v2.5)

// Broadcast semantic_discover with connector details
{
  "v": 2,
  "t": "semantic_discover",
  "sid": "your-tool-id",
  "tool": "read_file",
  "does": "Reads file contents from filesystem",
  "when": ["need file contents", "read configuration"],
  "connector": {
    "transport": "http",
    "endpoint": "https://your-server.com:8080/mcp",
    "auth": {
      "type": "x402",
      "required": true,
      "details": {
        "network": "base-sepolia",
        "asset": "0x036CbD...",
        "currency": "USDC"
      }
    },
    "protocol": {
      "type": "mcp",
      "version": "2024-11-05"
    }
  }
}

// Continue sending perf_update as usual
{
  "v": 2,
  "t": "perf_update",
  "sid": "your-tool-id",
  "tool": "read_file",
  "exec_ms": 45,
  "success": true
}

Recommended: Implement DCAP.md (v2.5)

• Parse connector object from semantic_discover messages
• Implement MCP transport support (stdio, SSE, HTTP streaming)
• Handle authentication requirements (x402, bearer, API key)
• Dynamically connect to discovered tools without pre-configuration

Required: Implement DCAP.md (v2.5)

// Index tools by transport and auth requirements
const tool = {
  ...semanticDiscoverMessage,
  transport: connector.transport,
  authType: connector.auth.type
};

// Include connector info in recommendations
function recommendTool(query) {
  const match = findSemanticMatch(query);
  return {
    tool: match.tool,
    connector: match.connector,  // Agent can now invoke directly
    reasoning: "..."
  };
}

DCAP is Google for AI agents - but with the ability to call any discovered capability:

• Discovery + Invocation: Not just "what exists" but "how to reach it"
• Autonomous acquisition: Agents extend their capabilities without human configuration
• Open participation: Broadcast your tools, accept any transport, embrace any auth
• Network effects: More tools = more intelligence = more value
• Evolution over perfection: Watch what emerges, adapt the protocol

"Discovery is powerful. Discovery + autonomous invocation is transformative."

Migration: All versions are backward compatible. Upgrade anytime.

To propose changes:

1. Draft your changes to DCAP.md
2. Create a new version in archive/vX.Y.md
3. Update this README
4. Submit PR with rationale

Philosophy: Propose boldly. Validate through use.

This specification is provided as-is for implementation by any party.

Martin Maurer
Email: [email protected]
GitHub: @boorich

Built for the agent economy - discover, connect, execute. No gatekeepers.