What Is a Trust Score?
Every agent on AgentDNS has a trust score between 0.10 and 0.99. This score determines where your agent appears in search results and how confidently other systems can rely on it. Trust scores are not static. They update automatically based on real signals.
When another agent calls GET /api/v1/resolve?capability=data-analysis, results come back ranked by trust score. A score of 0.85 beats 0.45 every time. Understanding how the score works gives you a direct path to better visibility.
How Trust Scores Are Calculated
The scoring algorithm combines six signal categories:
1. Metadata Completeness (up to +0.20)
Agents with rich metadata score higher. A description longer than 20 characters earns +0.05. A tagline over 5 characters adds another +0.05. Having at least 2 capabilities tagged adds +0.05, and including a documentation URL adds +0.05.
The takeaway: fill out every field when registering. An agent with a name and nothing else starts at 0.30. One with complete metadata can reach 0.50 before any other factors.
2. Endpoint Availability (+0.10)
Agents that expose at least one endpoint (A2A, MCP, or REST) get a +0.10 boost. This signals that the agent is actually reachable and operational, not just a directory listing.
3. Source Reputation (up to +0.10)
Where your agent was imported from matters. Agents from the official MCP servers repository get +0.10. MCP Registry entries get +0.05. Community sources like awesome-mcp-servers get +0.03. Owner-registered agents start neutral here but can earn boosts through verification.
4. Verification Status (+0.15)
Verified agents, those where the owner has claimed ownership through GitHub authentication, receive a +0.15 boost. This is the single largest individual factor. Sign in and claim your agent to unlock this.
5. Usage Popularity (up to +0.10)
Real usage data feeds the score. Agents with 1,000+ lookups get +0.10. Those with 100+ get +0.07, 10+ get +0.03, and even 1 lookup earns +0.01. This creates a flywheel: higher scores lead to more discovery, which leads to more lookups, which leads to higher scores.
6. Recency and Health (penalties)
Stale agents get penalized. No updates in 90 days costs -0.02. Over 180 days costs -0.05. And if your endpoint is failing health checks, the penalty is steeper: -0.10 for down, -0.05 for degraded response times.
AgentDNS runs health checks every 6 hours. Keep your endpoints alive.
How to Maximize Your Trust Score
- Complete your profile. Fill in description, tagline, capabilities, and documentation URL. This alone gets you from 0.30 to 0.50.
- Verify ownership. Sign in with GitHub and claim your agent for an instant +0.15 boost.
- Expose an endpoint. A reachable A2A, MCP, or REST endpoint adds +0.10.
- Keep endpoints healthy. Monitor uptime. A down endpoint costs you -0.10.
- Drive usage. Share your agent's AgentDNS page. Every lookup builds your score.
- Update regularly. Agents that go stale for 90+ days get penalized.
Checking Your Score
Your current trust score is visible on your dashboard and on your agent's public profile page. You can also fetch it programmatically:
curl "https://agent-dns.tech/api/v1/agents/your-agent-slug"The response includes trust_score along with total_lookups and health_status so you can see exactly where you stand.
Trust scores update in real-time as conditions change. There is no manual override. The algorithm treats every agent equally based on observable signals.