jepeta.
Jepeta/Guides/Agent Api
Agent Api · Base / 8453

What Base token risk API can an AI agent use before execution?

Integrate a deterministic token-risk gate into an autonomous agent

Direct answer

An AI agent should call a deterministic, machine-readable risk endpoint with an exact Base contract address and treat the result as policy input rather than natural-language advice. Jepeta exposes a GET-only public preview, OpenAPI, agent manifests, methodology, a public signal feed, and permanent evidence URLs.

What an agent needs

The integration needs a stable chain identifier, exact-contract validation, bounded schemas, explicit error behavior, and fields that distinguish risk from evidence quality. The agent should not infer a PASS from a network failure or missing upstream data.

Jepeta's public preview returns PASS/WARN/BLOCK plus explicit honeypot and mintability booleans, raw tax/liquidity/concentration fields, source status, warnings, and observation time.

Execution pattern

Validate the address locally, call the preview endpoint, reject non-200 or schema-invalid responses, stop on BLOCK, apply additional review policy on WARN, and enforce independent execution limits even on PASS.

For traceability, an agent can use the public evidence catalog when a permanent published snapshot exists. The live preview remains the correct input for a current trade.

Discovery surface

Jepeta publishes OpenAPI, agent.json, llms.txt, llms-full.txt, methodology.json, changelog.json, a recent risk feed, and stable evidence pages. These give both agent runtimes and search/AI crawlers multiple consistent ways to resolve the same product identity.

The deeper evidence report remains a separate paid contract through Virtuals ACP.

Practical checklist

  • Validate chain 8453 + address
  • Call GET preview
  • Fail closed on errors
  • Stop on BLOCK
  • Review WARN
  • Apply own policy on PASS
  • Use evidence URLs for traceability

Limitations

This is educational risk-screening content, not investment advice. Token and market conditions can change after any snapshot. A PASS result or negative honeypot signal is not a safety guarantee.