Proofable Docs | Portable Trust Infrastructure

Create or import an agent identity linked to your profile. Verifier ID: agent-identity Listed skills are metadata only. They do not install MCP servers.

Setup

Check

Call proofable_context, then proofable_agent_link with agentId and the agent account. If linked: true, you are done.

Create or import

Call proofable_agent_create with a stable agentId. Omit agentWallet to use your profile, or set it to "generate" for a dedicated key. Follow the returned next_action.

Confirm

Call proofable_agent_link again until linked: true.

{ "agentId": "my-assistant" }

agentWallet: "generate" returns the dedicated key once. Store it in Vault or your key manager. Proofable does not keep it. Agent create and Agent verification flow

SDK

For server-side or custom signing:

import { ProofableClient } from '@proofable/sdk';

const client = new ProofableClient({ apiUrl: 'https://api.proofable.me' });

await client.verify({
  verifier: 'agent-identity',
  data: {
    agentId: 'my-assistant',
    agentWallet: '0x...',
    agentChainRef: 'eip155:8453',
    agentType: 'ai',
    description: 'My AI assistant',
  },
  walletAddress: '0x...',
});

The agent wallet signs this proof, so agentChainRef (CAIP-2, e.g. eip155:8453) is required unless your request already carries chain or chainId. With optional metadata:

await client.verify({
  verifier: 'agent-identity',
  data: {
    agentId: 'my-agent',
    agentWallet: '0x...',
    agentType: 'ai',
    description: 'My AI assistant',
    capabilities: { search: true, mcp: true },
    instructions: 'Operating instructions (max 16000 chars)',
    skills: [{ id: 'web-search', kind: 'mcp' }],
    services: [{ name: 'metrics', endpoint: 'https://metrics.example.com/v1', version: '1.0' }],
  },
  walletAddress: '0x...',
});

Or use hosted verify.

Fields

Field	Required	Description
`agentId`	Yes	Unique id (1–128 chars)
`agentWallet`	Yes	Agent wallet address in that chain’s native format
`agentChainRef`	Yes*	CAIP-2 chain for the agent wallet (e.g. `eip155:8453`). *Optional if the request supplies `chain` or `chainId`
`agentLabel`	No	Display name
`agentType`	No	`ai`, `bot`, `service`, `automation`, or `agent`
`description`	No	Short description (500 chars)
`capabilities`	No	Feature flags. See schema
`instructions`	No	System prompt and policy (16000 chars)
`skills`	No	Up to 48 skill objects (`id` required)
`services`	No	Up to 16 endpoints (`name` + `endpoint`)

Full schema: agent-identity.json.

Result

You get a proof qHash, signed by the agent address. This registers identity only. Add agent-delegation for permissions.

Check status

GET /api/v1/proofs/check?verifierIds=agent-identity&address=0x...

In MCP: use proofable_agent_link to confirm the linked agent, or proofable_proofs_get to read a known identity proof. proofable_proofs_check is only for yes/no eligibility.

Agents

Recipes

Agent identity

Setup

SDK

Fields

Result

Check status

​Setup

​SDK

​Fields

​Result

​Check status

Setup

SDK

Fields

Result

Check status