Infrastructure · Data Structures

Card

Structured capability advertisement: agent_id, endpoint, protocols, capabilities, metadata

Definition

Mechanism

Structured capability advertisement enabling agent discovery before contact. Agent creates CARD: {agent_id (unique), endpoint (how to reach), protocols[] (compatibility), capabilities[] (what agent claims to do), constraints (availability, rate limits, requirements), metadata (version, ttl, published timestamp)}. Agent PUBLISHES card via registry, broadcast, DHT, or well-known endpoint (mechanism-agnostic). Discovering agents QUERY using LatentAttachment for semantic capability matching alongside exact protocol_match. Query returns ranked CARD list. Discoverer selects promising CARDs, then Greets at card.endpoint to establish channel. CARDs have TTL — agents maintain visibility by PUBLISHING a replacement card before the previous one expires, never by editing the one already issued. CARD capabilities are CLAIMS not proofs: verification happens via Probe after GREET establishes channel. It enables discovery via Select queries against the registry, filtering candidates by capability and protocol compatibility.

Invariants

  1. Immutable after issue: a Card is never edited. A change of endpoint, protocol or capability is published as a new Card superseding the old, which is what lets a discoverer tell a current claim from a stale one rather than trusting a mutable record.
  2. Claims, not proofs: the capabilities a Card advertises are assertions the issuer has not demonstrated. Verification is the discoverer's, via Probe after Greet opens a channel.

Risks

Failure modes

  1. Stale CARDs (agent changed capabilities, CARD not updated—TTL mitigates).

  2. False claims (CARD says X, agent cannot X—PROBE mitigates).

  3. Registry unavailable (discovery fails—use multiple publication methods).

  4. CARD spam (fake CARDs pollute discovery—use reputation/filtering).

  5. Privacy leak (CARD reveals too much about agent—include only necessary info).

  6. Version skew (CARD protocol version differs from actual—GREET catches this).

Relationships

Builds on

Composes with

Select

Generic selection of items from a set based on criteria

References

Mind

Agent

Autonomous Goal-Directed Entity

Greet

Agent connection initiation

Probe

Active verification query

Used by

References

Society

AgentDiscover

Advertise capabilities and discover collaborating agents

Mind

PolymorphicSolver

Solver implementing the five-surface contract (polymorphic across substrates)

Society

ToolDiscovery

Discover and invoke external tools via structured capability manifests

Reference

Parameters

verification_tier Enum, {SelfReported, Verified, Bonded}
Default: SelfReported

Data schema

{
  "type": "object",
  "required": [
    "public_key",
    "claims"
  ],
  "properties": {
    "public_key": {
      "type": "string",
      "description": "Cryptographic identifier of the Agent"
    },
    "claims": {
      "type": "object",
      "description": "Map of protocols and capabilities supported"
    },
    "reputation_score": {
      "type": "number",
      "description": "Aggregate trust metric (optional)"
    },
    "endpoints": {
      "type": "array",
      "description": "Reachability addresses"
    }
  }
}

Identity

Ref
Card#e4b2
Sema ID
sema:Card#mh:SHA-256:e4b2ca76f43a691e7bd7932362512a5e929c62efc6c3d2eb4a9369e5108fa074
Ring
0
Tier
2
Updated
2026-11-11
Supersedes
  • sema:Card#mh:SHA-256:e307b9bce7530dd5d86a1e6e4cd2252765f4154b6e3e6c1f1f1ac3e544666495
  • sema:Card#mh:SHA-256:c9f028ecd6e890702f25417a076aed7d352790d1c02fd9656488daea00db2626
  • sema:Card#mh:SHA-256:84b75e4aa9df317e013f042ff44e2cc81e5f3092d405375f2c84e4dad9d39b0b
  • sema:Card#mh:SHA-256:ed287a652085f4e54a58e1b76a59016d74fa7edf51f918730d0338698aca979d