About Agentic Card

Agentic Card is a protocol-agnostic registry for AI agents. We crawl, normalise, and evaluate agents across multiple protocols so developers and organisations can discover and trust the right agent for the job.

Trust Scoring Methodology

Every agent receives a trust score from 0–100 based on six weighted dimensions. The score is recalculated regularly as new data comes in.

Liveness

20%

Checks whether the agent has a live, reachable endpoint and measures uptime over 30 days of monitoring.

Completeness

15%

Evaluates how thoroughly the agent card is filled out — name, description, version, owner info, tools, modalities, and more.

Ownership

20%

Measures identity verification from unverified through email-matched to fully verified domain ownership.

Community

15%

Assesses open-source engagement: GitHub stars, forks, and follower counts normalised against category benchmarks.

Maintenance

20%

Examines recent commit activity, release cadence, contributor diversity, and whether the project is archived.

Security

10%

Verifies HTTPS, valid TLS, minimum TLS 1.2, authentication methods, and presence of an open-source licence.

Supported Protocols

We index agents from multiple protocols and normalise them into a common format for easy comparison.

A2A (Agent-to-Agent)

Google's open protocol for AI agent interoperability. Agents publish JSON cards describing capabilities, authentication, and supported modalities.

Wildcard

An emerging standard for agent discovery and capability advertisement, including /.well-known/agents.json and OpenAPI documents we fetch from well-known paths and site roots. Focuses on transparency, pricing information, and human-readable agent metadata.

MCP (Model Context Protocol)

Anthropic's protocol for connecting language models with tools, data sources, and external services. We index MCP server manifests as agents.

llms.txt

Plain-text files at /llms.txt that describe how LLMs should use a site or product. We ingest them and map the content into the same registry schema as structured agent cards.

How It Works

From submission to live listing in five steps.

Crawl

Fetch agent card from URL or raw JSON

Normalise

Map protocol-specific fields to a common schema

Enrich

Pull GitHub stats, verify TLS, check DNS

Evaluate

Score six trust dimensions with weighted formula

Display

List the agent with trust badge and health monitoring