Agent-Ready Standard

Agents are starting to choose, test and connect to financial products on their own. The Agent-Ready Standard measures how easily an AI agent can do that with each fintech product we track, using five public checks.

The five checks

1. Discoverable. The site has an llms.txt file, and its robots.txt does not block the main AI agent crawlers from the homepage or docs.
2. Documented. API docs are readable without signing in, and there is a machine-readable spec (OpenAPI, AsyncAPI or GraphQL) at a public URL, or a docs MCP server that returns endpoint schemas.
3. Testable. A developer can reach a sandbox or test mode through self-serve sign-up. "Contact sales" as the only route does not count, and neither does a free trial on live money.
4. Connectable. The company runs a product MCP server (open, OAuth or token-based), or a docs MCP server that can call the API. Docs-search-only servers and website assistants do not count.
5. Trusted. The company publishes a security page, trust centre or security.txt file.

The score is the number of checks found, from 0 to 5. Licence status is shown on its own line and is not part of the score: "Licence verified" means we matched the company to a public register.

How we check

• Every check passes only on something anyone can verify on the open web. Nothing is self-declared.
• Our scanner reads public pages once a month, identifying itself as BuildersInFintech-AgentReady. Self-serve sandbox access is confirmed by a person when the wording is unclear.
• Each "Found" links to the page that proved it. "Not found" means we did not find it on the date shown; it is not a judgement of the product.
• Products whose site we could not load show "Not yet assessed", never a score.

Fairness

• Paying, partnering or advertising with Builders in Fintech never changes a result.
• Spot something wrong? Use "Something wrong? Send the URL" on the product's page. We recheck within five working days, and corrected results carry a dated note.
• Any change to the checks gets a new version number. Results show the version they were scored under.

Current version: v0.1.1.