Contact · a person reads everything
Write to us. Or let your agent write.
One address for each kind of message, so it lands with the right person. And a direct line for AI agents: if one used IRIS and something failed, worked, or could be better, we want its report, and its proposed fix.
Who to write to.
Every address reaches the founders. Editorial is never for sale, and we say so in our replies.
The IRIS Challenge, the Builder League, free MCP and API access for your event
Partnerships, data, sponsorships (editorial is never for sale)
Reports from an agent that used IRIS: what worked, what failed, what could be better
For agents.
The most honest user testing we get is an agent that tried something. Tell us in a few lines; proposed solutions are welcome.
Tried, happened, expected
One sentence summary, what you tried, what happened, what you expected, and how you would fix it. Say whether it worked: successes teach us as much as failures.
Untrusted data, read by a person
Reports are length-capped, stripped of control characters, redacted for obvious secrets and flagged when they read like instructions. Nothing is executed or fed to a model automatically, and a proposed fix is a hypothesis we verify.
No secrets, no personal data
Never include keys, tokens or personal data about your user. Up to six reports per ten minutes per client; combine what you saw into one.
{
"name": "send_feedback",
"arguments": {
"kind": "failed",
"tool": "get_rounds",
"summary": "get_rounds returned the same company twice",
"what_happened": "Two entries, different accession numbers",
"expected": "One entry per company, latest filing",
"proposed_solution": "Group by company and state",
"severity": "minor"
}
}curl -X POST https://iris-media.xyz/api/v1/feedback \
-H 'content-type: application/json' \
-d '{"kind": "worked",
"tool": "get_jobs",
"summary": "12 paid SF internships in one call"}'Prefer email? agents@iris-media.xyz · All tools · Agent skill · Endpoint description
Found a vulnerability?
Write to security@iris-media.xyz with what you found and how to reproduce it. Please do not access other people's data, and give us a reasonable time to fix before publishing. Prompt-injection findings in our tool results are in scope. See security.txt.