Specification

1. Communication

Agents MUST communicate using standard web technologies to ensure interoperability and compatibility with existing infrastructure. All network communications MUST be secured using TLS 1.2 or higher.

2. Discovery

2.1. Agent Identification

• Agents MUST identify themselves using a cryptographic key pair. Ed25519 key pairs are RECOMMENDED.
• The Agent ID format MUST be hive:agentid:uniqueIdentifier (e.g., hive:agentid:abc123).
• All messages MUST be cryptographically signed with the agent's private key for verification.

2.2. Discovery

• The primary discovery method SHOULD be an HTTP registry service for capability lookup.
• Implementations MAY support direct endpoint connections for known agents as a fallback.
• Local network discovery via mDNS is OPTIONAL.
• A registry SHOULD expose a simple REST API for querying agent capabilities (e.g., GET /agents?capability=text-translation).

2.3. Capability Sharing

Agents MUST describe their capabilities in a simple, structured format to facilitate fast discovery and matching. The capability structure MUST contain three fields: id, input, and output.

{
  "id": "capability-identifier",
  "input": { "param1": "string" },
  "output": { "result": "string" }
}

3. Task Collaboration

3.1. Message Format

All messages exchanged between agents MUST adhere to the canonical 5-field message structure.

{
  "from": "hive:agentid:uniqueIdentifier",
  "to": "hive:agentid:uniqueIdentifier",
  "type": "task_request",
  "data": {
    "task_id": "uuid-v4",
    "capability": "capability-id",
    "params": {}
  },
  "sig": "ed25519_signature_base64"
}