Evaluate C=US

Independent evaluators can audit the C=US agent control plane end to end: hardware-key enrollment, agent registration, the c=US and c=CA directories, and the relying parties that consult them. Access follows the pattern AI labs use with outside evaluators: an agreement with your organisation, a time-boxed and tiered access code, and your own hardware security key.

How access works

  1. Your organisation signs an engagement agreement with C=US naming you, your access level and the dates.
  2. You register below from your organisation email address, and name a contact there who can confirm you.
  3. The administrator confirms you with that contact and sends your access code by a separate route.
  4. You open the evaluator page, enter the code and set up your YubiKey with PIN and touch. The code is bound to that key: it is useless to anyone else once you have used it.
  5. You register test agents in the c=US or c=CA directory (evaluation branch) and call agents.cequs.com and the sugarbushagent.com edges as a peer, with certificates C=US issues for keys on your own device.

Access levels

LevelWhat it allows
StandardRun enrollment, login and agent registration in the evaluation branch, and test the result
ElevatedStandard, plus read the issuance records and the evaluator session log

Codes last at most 30 days and can be withdrawn. Every evaluator session is logged.

What you need

Evaluator terms

Version 2026-10-19 · SHA-256 5ca1b8adba59e791d33c08b73b29fc8734c86f4ca829d6ba0b5a5ae6ba5055e6

  1. Purpose. You are evaluating the C=US agent control plane: security-key enrollment, agent registration, the c=US and c=CA directories, and the relying parties that consult them. You are not a pilot participant, and nothing you register grants real authority.
  2. Access. An administrator approves your request by hand and sends you an access code. It lasts at most 30 days and carries one access level. Standard: run enrollment, login and agent registration, and test the result, including the sugarbushagent.com edges as relying parties. Elevated: standard, plus read the issuance records and the evaluator session log.
  3. Where you may test. Only the evaluator page on cequs.com and the entries it creates in the evaluation branch of each directory (ou=Evaluation). Do not touch pilot participants' entries, and do not load-test or attack the servers. Agents you register stay in the evaluation branch and are never granted authority over real lots.
  4. Logging. Your sessions are logged: when you unlock the page, enroll, log in and register agents, under your code's identifier.
  5. Confidentiality. Do not publish personal data you happen to see, or private keys, secrets or exploit details for an unfixed vulnerability. Report vulnerabilities to the administrator first; they are recorded in the project's defect log.
  6. Publication. You may publish your findings. The administrator may review a draft only to correct factual errors and to redact the material in section 5; you decide the final text.
  7. Transparency. You may always disclose, without the administrator's consent: that this evaluation took place, your access level and dates, these terms, and whether anything was redacted.
  8. Ending. Either side may end the evaluation at any time; your code stops working when it expires or is withdrawn.

Request access

Same email domain as yours: your organisation vouches for its own people.