About Attify
Attify independently audits what AI agents actually achieve, using real data from the companies that use them. Each job an AI agent does gets a published standard that asks three questions of every claimed outcome: did it happen, did it stick, did a human have to redo it.
What Attify publishes#
- Standards. One per job type, with the rules, the judge prompt word for word, and a changelog: customer-support resolution, meetings booked by AI sales agents, and appointments booked by AI receptionists. Coding tasks, collections and back-office processing will follow as standards of their own.
- Audited scores. A vendor's audited rate, measured on its independent customers' own systems, with a live badge and a signed attestation anyone can check at /verify. Listed in the registry.
- Dashboards for buyers. Each month, for each AI agent a company pays for: what it claimed, what passed the audit, and what each audited outcome cost.
- The Attify Index. Audited outcomes by job type, pooled from buyers who choose to contribute anonymous data, and never naming a company. It publishes only when enough companies contribute.
How it works#
Attify reads the customer's own systems through the customer's own authorization: help desks (Intercom, Zendesk, HubSpot, Salesforce, Gorgias, Freshdesk), CRMs and calendars (HubSpot, Salesforce, Google Calendar, Microsoft 365, Calendly, Cal.com), agent platforms (Vapi, Retell AI, ElevenLabs, Bland AI, Voiceflow) and observability tools (Langfuse, LangSmith). Any other agent can stream its interactions as they happen through the Ingest API, the Python and TypeScript SDKs, or OpenTelemetry. The full method is on /methodology.
Principles#
- Answer first. Every result opens with one sentence written from the numbers by a fixed template, then the claimed rate, the audited rate and the gap.
- The fairest version of the vendor's own rule. If a skeptic could shrink a gap by pointing at something no careful vendor would count, the standard changes.
- Every number states its evidence. Its tier, its provenance, its sample size and its 95% interval.
- Neutral. Attify describes where outcomes fail and never advises a vendor on raising its score. No price depends on a result. See Is Attify independent?
- Agent-readable. Every result exists as typed JSON with a versioned schema, through the API and an MCP server, so other agents can use it without docs.
Who runs it#
Attify is operated by Dor Tagger. Attify (attify.app) is not affiliated with attify.com. It was founded in 2026 by Dor Tagger; Dor Tagger's other businesses, and the recusal rule that keeps them off Attify's registry, are on Independence.
Contact#
- Standards comments: see How to comment
- Corrections: corrections policy
- Vendors' right of reply: right of reply
- Press: press
- Security: /security
Questions
What does Attify do?
- Attify checks the outcomes an AI agent claims (resolved conversations, booked meetings, booked appointments) against the customer's own records, under a published standard for each job type, and reports the audited rate next to the claimed one.
Who is Attify for?
- Companies that pay for AI agents and want to know whether they work, and companies that build AI agents and want their numbers believed. Developers connect agents through the Ingest API, the SDKs or OpenTelemetry.
Which AI agents can Attify verify?
- Any agent doing a job type with a published standard: customer support, sales meetings and appointment bookings today. It reads six help desks, six CRMs, calendars and scheduling tools, five agent platforms and two observability tools, and takes any other agent through the Ingest API.
Is Attify a review site?
- No. It publishes no opinions or star ratings. It publishes measured outcomes under published rules, with the evidence tier and provenance of every number.