An autonomous DevOps agent that operates without human supervision.
Mendral's AI engineer runs entirely inside Blaxel sandboxes, examining code and executing commands across customer repositories with hardware-level isolation and no human in the loop.
- [INDUSTRY]
- Vertical AI product
- [PRODUCTS]
- SandboxesProxy routingStatic IPs
[THE STORY]
How Mendral runs on Blaxel
The problem
An autonomous DevOps agent executes real commands against real customer repositories. Without a human reviewing each step, the runtime itself has to guarantee that a mistake, or a prompt injection, cannot escape.
What changed
Every Mendral run happens inside a hardware-isolated Blaxel microVM. Credentials are injected by the egress proxy rather than stored in the sandbox, and each run is auditable.
At scale
Mendral's AI engineer operates across customer repositories around the clock, with no human in the loop.
[RESULTS]
What changed
- Isolation
Hardware-level
One microVM per run, no shared kernel between customers.
- Credentials
Never in the sandbox
Secrets injected in flight by the managed proxy.
- Supervision
None required
The agent runs unattended, with every request logged.
