API

Read-only JSON, no key required, one company per request. Requests are rate limited per client (HTTP 429 with Retry-After when exceeded). There is no bulk endpoint; bulk access, history and change alerts are planned as a keyed tier.

Endpoint

GET /api/v1/companies/{id}

One company with surfaces, every claim (tier, quote, source, date), per-agent results, robots.txt checks and change events.

Example

curl "https://agentaccepted.com/api/v1/companies/amazon"

Reading a company

• overall: allowed, conditional, blocked, selective, conflicting or none. none means no policy found, not allowed.
• stated and technical: what the company says and what its robots.txt does, kept separate. stated_confidence is confirmed, reported or lead.
• agents[]: one entry per product with status (null means no signal), via (robots or statement) and advisory (the vendor says the agent may ignore robots.txt).
• claims[]: tier T1 to T4, quote, evidence_url, observed_at.
• why: one line explaining the verdict.
• coverage: how many of the agents robots.txt can speak to have a result (covered of total). A verdict is only about the agents it covers.
• check_issue: set when there is no verdict because our last check was refused or unreachable (kind, status, host). policy=unchecked lists these companies.

See the methodology for how verdicts are derived. Machine-readable overview: llms.txt.