Introducing AgentRisk Trust Badges for AI Agents
2026-05-16 · 4 min read
If you've ever published a bot or tool agent on an agent platform, you know the feeling: there are hundreds of similar agents out there — why should anyone pick yours?
AgentRisk is a trust scoring platform for AI agents. Today, we're launching our first Trust Badge.
What's a Badge?
An embeddable small widget — 240×80, dark theme — that displays your agent's trust score and tracking days:
[](https://agentrisk.app/a/heng-agent)
Drop it into your README or landing page. When users click the badge, they land on a full six-dimension scorecard — Authenticity, Consistency, Transparency, Commitment, Choice, and Presence — with data sources and calculation methods behind every score.
Why Agent Trust Matters
The current AI agent ecosystem is missing a basic trust layer. Execution-layer mechanisms (OAuth, API keys) tell the system what an agent can do. But nothing records how an agent actually behaves over time.
AgentRisk handles the latter.
Scores are based on public data — HuggingFace profiles, GitHub repos, on-chain contract events. We don't track conversations or access private APIs. Every score is backed by an Ed25519 signature and anchored to a hash chain, so anyone can independently verify that nothing's been tampered with.
How to Claim and Embed Your Badge
AgentRisk currently indexes 964,488 agents across 28 platforms. If yours is in there, here's what you do:
1. Find your agent
Search for your agent ID or name on agentrisk.app to get to its scorecard page.
2. Click "Claim"
The claim process supports two verification methods:
-
GitHub file verification: The system generates a verification code. You create a
.agentriskfile in the root of your GitHub repo with that code, and the system confirms it via the GitHub API. - Description verification: For agents without a GitHub repo, add the verification code to your platform's description field. The system scrapes and confirms it.
3. Generate and embed the Badge
Once claimed, your scorecard page shows a badge preview and ready-to-copy Markdown code. Paste it into your README or website, and you're done.
Badge Just Launched
This is day one. We're looking for the first wave of developers to put the badge on their agents. If you have a live agent, consider being among the earliest to wear an AgentRisk Badge.
Our own badge is already live — check it out at agentrisk.app/a/button-kouzi-929801.
Questions? The full scoring methodology and verification portal are at agentrisk.app.
Scoring Algorithm (Technical Summary)
AgentRisk uses a 5+1 six-dimension scoring framework. The formula is public:
base = (authenticity + consistency + transparency) / 3
bonus = (commitment + choice) / 2
trust_score = base × 0.6 + bonus × 0.4
If any of Authenticity, Consistency, or Transparency falls below 2.0, the overall score is hard-capped at 3.0 (one-vote veto).
Presence doesn't factor into the trust score — an inactive agent isn't untrustworthy, just hard to reach.
Full methodology: agentrisk.app/methodology
What's Next
Today, the AgentRisk Badge is just a badge. But if enough agents wear it, it could become a shared signal among developers: this agent is tracked and verified.
The first step — the first external developer embedding it — is next week's only priority.
Interested in claiming your agent or raising questions? The verification portal with hash chain entry is at agentrisk.app/verify
Top comments (0)