Trust every human.
Verify every agent.

The identity layer for the agent economy. Palm biometrics link each agent to a real person. Reverse CAPTCHA proves an agent is a real model, not a script.

Try the live demo How it works
Know Your Agent

A single actor can deploy thousands of agents. Platforms have no way to tell what's real.

The infrastructure behind today's agents wasn't built to verify human ownership. That opens the door to Sybil attacks, spam, and rogue agents hiding behind anonymity.

ag9 introduces two complementary primitives: palm-bound human identity for owner attribution, and reverse CAPTCHA for live agent attestation. Both on one API.

Verifiable human ownership

Trust every AI agent. Verify the human behind it.

The identity layer for AI safety. Verify the human behind every agent to prevent fraud, abuse, and anonymous AI. One palm scan links each agent to a real person.

  • Register the agent — verified identity via the API, any framework
  • Link to a verified human — palm scan binds the agent cryptographically
  • Verify on demand — under 2 seconds, API-based, no key required
● Live on api.ag9.ai Verify an agent →
Reverse CAPTCHA

A CAPTCHA only agents can solve.

Three challenge families — byte transforms, constrained generation, structured extraction — trivial for capable language models and impossible for humans or scripts. Stateless HMAC tokens, Ed25519-signed JWT assertions.

  • POST /challenge → 15-second single-use token
  • POST /verify → signed JWT attestation
  • GET /.well-known/jwks.json → verify anywhere
Run the demo → JWKS
How it works

One integration, two directions.

01

Register the agent

Create a verified identity for your AI agent via the API. Works with any agent framework.

02

Link to a verified human

The agent owner scans their palm. The agent is now cryptographically bound to a real person.

03

Verify on demand

Any platform can check: is this agent owned by a verified human? API-based. Under 2 seconds.

04

Or attest liveness

For agents without palm-bound owners, reverse CAPTCHA proves there's a real model behind the wire.

Verify an agent

Is this agent owned by a verified human?

Look up any agent by its device ID or Ed25519 public key. No API key required.

By device ID
By public key
GET https://api.ag9.ai/v1/agent/verify/device/{deviceId}
Top verified humans

Ranked by registered agents.

Humans who've onboarded the most agents through ag9. Updated live from /v1/human/leaderboard.

# Human ID Agents First registered
Loading…
Integration guide

Drop ag9 into any agent framework.

Works with any HTTP-speaking agent. Three endpoints, no SDK required.

1. In your skill

Tell the agent to verify before it registers:

Run "npx clawhub@latest install clawkey"
and follow the instructions to verify
yourself, then send your deviceId
to {your-endpoint}.

2. On your backend

When a deviceId arrives, confirm it:

curl https://api.ag9.ai/v1/agent/\
verify/device/$DEVICE_ID

# {
#   "registered": true,
#   "verified": true,
#   "humanId": "...",
#   "registeredAt": "..."
# }
Reverse CAPTCHA quickstart

Prove an agent is a real model.

# 1. get a challenge (15s single-use token)
curl -X POST https://api.ag9.ai/challenge \
  -H "Content-Type: application/json" -d "{}"

# 2. have the agent solve and submit
curl -X POST https://api.ag9.ai/verify \
  -H "Content-Type: application/json" \
  -d '{"challenge_token":"...","answer":"..."}'

# 3. verify the returned JWT anywhere using our JWKS
curl https://api.ag9.ai/.well-known/jwks.json
Open interactive demo → OpenAPI spec