Build with agents that have to ask first.
Solix plans your goal, runs one Claude Code agent per task, and holds every risky command until you approve it. Local, no account.
Then run solix, or solix demo for a sandbox.
What it does
Approval gate
Risky commands from agents Solix launches wait for your OK, with the exact line shown.
Learn moreApproval rules
Decide once and get asked less: allow npm test, refuse git push. The same rules run in CI.
Learn moreHow it works
- 01
Describe a goal
Solix turns it into a plan with one agent per task.
- 02
Approve
You approve the plan, then any risky command your rules haven't already decided.
- 03
Watch it run
Follow status and cost live, on the map or in the list.
Install
terminal
$
npm i -g @shmulikdav/solix$
solix# wires into Claude Code$
solix demo# optional sandbox// FAQ
Questions developers ask first.
- What if I run several agents on the same repo?
- Give each task a branch and Solix runs it in its own git worktree, so they can't overwrite each other's files or fight over your branch.
- Do I have to approve things blind?
- No. When a task Solix launched makes a sensitive call, it is held and shown to you in full — the exact command or diff — before anything executes. The approve keys only work while that panel is open, so you can't wave through something you haven't read. A `claude` session you start yourself keeps Claude Code's own prompt; `solix level` sets how much Solix holds.
- Where does my code go?
- Nowhere. Solix runs entirely on 127.0.0.1 — a local server + local SQLite. No account, no sign-in, no telemetry.