Myagi Open Agent Registry

What Myagi - Open Agent Registry does

Myagi - Open Agent Registry evaluates how accessible a website or service is to AI agents. It combines direct scans with data from authoritative registries and standards sources, then produces a score, a readiness label, and a checklist of suggested next steps.

The project covers six categories: Rentals & Travel, Shopping & Commerce, Content & News, Social & Community, Developer Platforms, and AI & ML Platforms. Its public deployment contains more than 1,400 tracked services after importing data from the llms-txt-hub community list. Each service has a live score and an approval-path checklist.

How it works

The scanner examines a target URL for signals that can help agents discover, authenticate with, or use the service. Checks include llms.txt, agents.txt or agents.json, AGENTS.md, the well-known agent-card file, robots.txt and content signals, OpenAPI descriptions, OAuth or OIDC discovery, RSL licensing information, and presence in MCP or A2A registries.

Scores use a published rubric rather than an opaque model. For example, an official MCP Registry entry and an A2A Agent Card each receive 25 points, while llms.txt receives 8 points. Blocking automation can reduce the score substantially. The resulting total maps to labels such as Agent-ready, Agent-friendly, Partially agent-accessible, API-only / limited, Unknown, or Likely blocked.

Community directories help identify domains to scan, but they do not automatically provide scoring credit. The project distinguishes those discovery sources from official registries and standards references.

Setup and configuration

The repository runs as a Next.js application with TypeScript and Tailwind CSS. Install dependencies with pnpm, start the development server, and open the local address:

pnpm install
pnpm dev

The initial version uses Next.js API routes for scanning and a local JSON cache. It does not require a separate backend service for the MVP. The application is designed for native deployment on Vercel; the README identifies Postgres or Vercel KV as options for production-scale storage, rather than current MVP requirements.

Tools and capabilities

Myagi - Open Agent Registry provides scanner and platform capabilities for working with its registry data. Supported functions described by the project include:

• Scan a URL for agent-access and discovery signals.
• Calculate a reproducible score from the published rubric.
• Assign a readiness label and industry category.
• Rank services on comparative category leaderboards.
• Generate approval-path checklists with links to developer applications, OAuth registration, partner programs, and standards-adoption steps when available.
• Monitor changes over time.
• Query registry information through a public API.

The project also aggregates upstream registry information instead of rescanning or replacing the work of those registries. Its categories and scores are intended to help agents decide what to inspect before calling an external service.

Limitations and notes

The repository describes the project as pre-alpha, although a public deployment is available. The scan quality depends on the parsers and the information published by each service. The README says the MVP parsers are intentionally conservative, so improving them may change data quality and scores.

A score is evidence of discovered signals, not a guarantee that an agent can complete a task on the target service. The rubric includes negative points for explicit bot prohibitions, blocked automation without an API alternative, and unparseable or unknown terms. Approval-path data is also an active contribution area, so checklist coverage may vary between services.

The public registry is based on an original seed plus imported community data. Community listings are used for domain discovery and are not themselves authoritative scoring evidence.