Found something? Tell us first.
We welcome reports from security researchers, pay a bounty for valid findings, and will not pursue legal action against anyone who follows this policy in good faith.
One address, encrypted if you like.
Please do not open a public GitHub issue for security problems.
/.well-known/security.txt.Timelines we hold ourselves to.
Acknowledge
A human confirms we received your report and who is handling it.
2 business daysTriage
We reproduce, assess severity, and tell you whether it is in scope and what we plan to do.
7 daysFix
Critical issues ship as fast as we can build them; others within the window. Fleet machines update automatically at night.
90 days maxDisclose
We publish an advisory and credit you, if you want credit. You are free to publish once a fix is out or the window has passed.
CoordinatedWe pay for valid reports.
Amounts are set at triage by the severity of what an attacker could actually do, not by the type of bug.
$100
Limited impact, needs unusual conditions. Information leaks with no sensitive data, minor policy inconsistencies.
$250
Affects one account with user interaction. Stored XSS in the console, CSRF on settings, auth logic flaws.
$500 – $1,000
Agent reaches the internet around the firewall, bypasses an Ask or Block rule, or reads credentials it should not have.
$2,000+
Cross-enclave access, reading another customer's data, decrypting backups without the key, or remote code execution on the control plane.
One payout per root cause; duplicates go to the first complete report. Paid by bank transfer or PayPal within 30 days of the fix shipping. Issues that need a stolen enclave passphrase, or that only affect out-of-scope components, are not eligible, but we will still say thank you.
What counts.
In scope
- controlclaw.com and the console
- ControlClaw Firewall (the open-source Rust project)
- Agent machine images and the update pipeline
- Enclave isolation, key handling, backup encryption
- Policy bypasses: an agent reaching the internet around the firewall
Out of scope
- OpenClaw itself; report upstream to that project
- Model providers and other third-party services
- Denial of service, rate limiting, spam
- Social engineering of staff or customers
- Findings from automated scanners without a working exploit
Safe harbor.
Test only against accounts you own or have permission to use. Do not access, modify or retain other customers' data; if you encounter it by accident, stop and tell us. Do not degrade the service for others. Give us a reasonable window to fix before publishing.
If you follow these rules, we consider your research authorised, will not pursue or support legal action against you, and will work with you if a third party does. Valid reports are paid under the bounty tiers above, and we say thank you publicly in our advisories and on the repository.
Questions about the policy itself: contact us.