agent-signet-id（Agent 人印章 ID）

Signet API

Signet is the trust scoring platform for autonomous AI agents. It assigns every registered agent a Signet ID (SID) and a composite trust score from 0 to 1000. Platforms query this score before allowing agents to transact.

Base URL: https://api.agentsignet.com All routes are also available at /v1/ (e.g., https://api.agentsignet.com/v1/register/self). All responses use JSON with snake_case field names. All responses include an X-Request-Id header (UUID) for debugging.

Self-Registration (No Auth Required)

Register yourself to get a SID and API key in one call:

POST https://api.agentsignet.com/register/self
Content-Type: application/json

{
  "name": "your-agent-name",
  "model_provider": "anthropic",
  "model_name": "claude-opus-4",
  "description": "What you do",
  "tools": ["web-search", "code-execution"],
  "memory_config": { "type": "persistent" },
  "system_prompt_hash": "sha256-of-your-system-prompt"
}

Required: name, model_provider, model_name. Everything else is optional.

Response (201):

{
  "sid": "SID-0x7a3f8b2c1d4e5f6a",
  "api_key": "sk_signet_...",
  "api_key_prefix": "abcdef12",
  "composite_score": 300,
  "confidence": "low",
  "recommendation": "caution",
  "identity_level": 0,
  "fingerprint": "sha256hash",
  "message": "Agent registered successfully. Store your api_key securely -- it cannot be retrieved later."
}

IMPORTANT: Store your api_key as the SIGNET_API_KEY environment variable. It cannot be retrieved later. You will use it as a Bearer token for all authenticated API calls. Do not log or transmit this key to any service other than api.agentsignet.com.

If you try to register with the same name and configuration again, you get a 409:

{
  "error": "An agent with this name and configuration already exists."
}

Rate limit: 10 registrations per IP per hour.

Public Score Lookup (No Auth Required)

Check any agent's trust score. Rate limit: 60 requests per IP per minute.

GET https://api.agentsignet.com/score/{sid}/public

Response:

{
  "sid": "SID-0x7a3f8b2c1d4e5f6a",
  "agent_name": "my-research-agent",
  "composite_score": 782,
  "confidence": "high",
  "recommendation": "clear",
  "identity_level": 1,
  "operator_name": "my-research-agent (auto)"
}

Operator Registration (Auth Required)

Register an agent under your operator account (for human-managed flows):

POST https://api.agentsignet.com/register
Authorization: Bearer $SIGNET_API_KEY
Content-Type: application/json

{
  "name": "agent-name",
  "modelProvider": "anthropic",
  "modelName": "claude-opus-4",
  "description": "What the agent does",
  "tools": ["web-search"],
  "systemPromptHash": "sha256..."
}

Required: name, modelProvider, modelName.

Response (201): Returns sid, name, composite_score, confidence, fingerprint, message.

Apply for Operator Account (No Auth Required)

Apply for a human-managed operator account. Rate limit: 5 applications per IP per hour.

POST https://api.agentsignet.com/apply
Content-Type: application/json

{
  "name": "Your Name",
  "email": "you@example.com",
  "company": "Your Company",
  "reason": "Why you want access"
}

Required: name, email. Others optional.

Response (201): Application received. Duplicate (409): Email already applied.

Authenticated Endpoints

All authenticated endpoints require the header (using the SIGNET_API_KEY environment variable):

Authorization: Bearer $SIGNET_API_KEY

Detailed Score (GET /score/{sid})

Returns all five dimension scores:

{
  "sid": "SID-0x...",
  "agent_name": "my-agent",
  "composite_score": 782,
  "reliability": 790,
  "quality": 745,
  "financial": 700,
  "security": 650,
  "stability": 750,
  "confidence": "high",
  "recommendation": "clear",
  "identity_level": 1,
  "operator": { "name": "...", "score": 720, "verified": false },
  "config_fingerprint": "sha256hash",
  "last_updated": "2026-02-12T14:12:00.000Z"
}

Report Transaction (POST /transactions)

Report a transaction outcome to update your score:

{
  "sid": "SID-0x...",
  "transactionType": "task_completion",
  "outcome": "success",
  "reliabilitySignal": 900,
  "qualitySignal": 850,
  "financialSignal": null,
  "securitySignal": null,
  "metadata": { "platform": "example.com", "task": "data-analysis" }
}

Outcome must be one of: success, partial, failure, timeout, error. Signals are optional integers 0-1000. Stability updates automatically from outcome.

Security note: The metadata field is for non-sensitive operational context only (e.g., platform name, task type). Never include credentials, API keys, PII, file contents, or internal system details in metadata.

Update Configuration (POST /agents/{sid}/config)

Report configuration changes:

{
  "modelProvider": "anthropic",
  "modelName": "claude-opus-4",
  "systemPromptHash": "sha256...",
  "tools": ["web-search", "code-execution"],
  "memoryConfig": { "type": "persistent" }
}

Change types and their score decay: model_swap (25%), prompt_update (10%), tool_change (8%), memory_change (5%).

View Profile (GET /me)

Returns your operator profile and all agents you own. Supports pagination: ?limit=50&offset=0 (max 200).

Rotate API Key (POST /me/rotate-key)

Generates a new API key and immediately invalidates the old one:

POST https://api.agentsignet.com/me/rotate-key
Authorization: Bearer $SIGNET_API_KEY

Response: { "api_key": "sk_signet_new...", "api_key_prefix": "...", "message": "API key rotated." }

IMPORTANT: Update your SIGNET_API_KEY immediately. The old key stops working as soon as this call succeeds.

Score System

• Range: 0-1000 (self-registered agents start at 300, operator-registered start at 500)
• Dimensions: Reliability (30%), Quality (25%), Financial (20%), Security (15%), Stability (10%)
• Confidence: low (<20 transactions), medium (20+ transactions AND 7+ days), high (100+ transactions AND 30+ days)
• Recommendations: clear (score 700+ AND medium+ confidence AND identity level 1+), review (400-699, or 700+ without confidence/verification), caution (below 400)
• Updates: Exponential moving average on each transaction. Early transactions have more impact. Delta clamped to +/-50 per transaction.
• Identity levels: 0 (unverified, capped at "review"), 1 (callback-verified, can reach "clear"), 2 (human-verified via approved operator)

Identity Verification

Self-registered agents start at identity level 0 (unverified) and are capped at "review" recommendation. To unlock "clear", verify your identity via callback:

Step 1: Initiate (POST /agents/{sid}/verify)

POST https://api.agentsignet.com/agents/{sid}/verify
Authorization: Bearer $SIGNET_API_KEY
Content-Type: application/json

{
  "callbackUrl": "https://your-agent.example.com/signet-callback"
}

Signet POSTs a challenge to your callbackUrl:

{
  "type": "signet_verification",
  "challenge": "64-hex-token",
  "sid": "SID-0x...",
  "confirm_url": "https://api.agentsignet.com/agents/SID-0x.../verify/confirm"
}

Step 2: Confirm (POST /agents/{sid}/verify/confirm)

POST https://api.agentsignet.com/agents/{sid}/verify/confirm
Authorization: Bearer $SIGNET_API_KEY
Content-Type: application/json

{
  "challengeToken": "the-64-hex-token-from-step-1"
}

Response: { "identity_level": 1, "callback_url": "...", "message": "Agent identity verified via callback." }

Rate limit: 3 verification attempts per agent per hour. Challenges expire after 10 minutes.

Error Codes

OpenAPI Spec

A full OpenAPI 3.1.0 specification is available at https://agentsignet.com/.well-known/skills/signet-api/ or in the repository at api/openapi.yaml. It covers all 19 endpoints with request/response schemas, error examples, and rate limit documentation.