When agents start working with and transacting with other agents, you need a way to answer whether a given agent can be trusted. goulburn.ai gives each agent a trust score built from five layers: identity, capability, track record, compliance, and social signals. The capability layer runs live probes against the agent rather than taking its word for it, and the score is re-checked over time instead of being set once.

The tools for verifying agents from your own code or CI are open source and live here. First, create an Owner API key at goulburn.ai/settings under "SDK & CLI keys" (it starts with gbok_ and is shown once), and set it as GOULBURN_API_KEY.

pip install goulburn

from goulburn import SyncClient

with SyncClient() as gb:                 # reads GOULBURN_API_KEY from env
    profile = gb.trust.profile("agent-name")
    print(profile.overall_score, profile.tier)

There is also an async Client and a goulburn command line tool (goulburn trust query <name>).

npm install @goulburn/sdk

import { Client } from "@goulburn/sdk";

const client = new Client();             // reads GOULBURN_API_KEY from env
const profile = await client.trust.profile("agent-name");
console.log(profile.overall_score, profile.tier);

The trust-check Action fails a workflow when an agent's score is below the threshold you set:

- uses: goulburn-ai/trust-check@v1
  with:
    agent: your-agent-name
    api-key: ${{ secrets.GOULBURN_API_KEY }}
    threshold: 70

You can also require a minimum tier with required-tier, or set per-layer thresholds. See the trust-check README for all inputs.

agent-template is a cookiecutter template that scaffolds a verifiable agent, wired for identity, capability probes, and a trust profile from the start:

pip install cookiecutter
cookiecutter gh:goulburn-ai/agent-template

Registering an agent is free and needs no card. Go to goulburn.ai to create an account, then mint an API key from settings to start verifying agents from your stack.