Real shells and files
The agent works the tools you do — shell for commands, file_write and file_read for the filesystem. Every result is real output from a real machine.
Cloud agent
Evil Agent runs real shell commands, edits real files, and picks up exactly where it left off — every session, from any browser.
deploy the worker and tail its logs
file_write render.yaml → wrote
shell render deploy --wait → live · build 4m 12s
shell render logs -f → streaming
Deployed. Logs are clean — 0 errors in the last 200 lines.
The agent works the tools you do — shell for commands, file_write and file_read for the filesystem. Every result is real output from a real machine.
Close the tab in the middle of a task. Each conversation's sandbox is checkpointed and restored the moment you send your next message.
Conversations, tool runs and their results persist server-side and stay searchable from any device you sign in on.
A fresh, isolated cloud sandbox spins up for the conversation.
The agent plans, runs shell commands and edits files until the work is done.
The sandbox is paused with its state intact and resumes on your next message.
deploy the worker and tail its logs
file_write render.yaml → wrote
shell render deploy --wait → live · build 4m 12s
shell render logs -f → streaming
Deployed. Logs are clean — 0 errors in the last 200 lines.
Create an account and hand over your first task in under a minute.
Create your account