Assistants with a role
Planner, Team leader, Code reviewer, Slide designer, Data and Excel analyst, Researcher. Press one and a new chat opens in that role with a first request in the composer.
Introduction
Hand it over. It gets finished.
Give the job to an assistant with a role and approve only the risky commands. The LEVEL11 agent, Claude Code and Codex work in the same workspace.
18 models on one account
Claude, GPT, Gemini, Grok, DeepSeek, GLM, Qwen. Leave it on auto and the LEVEL11 Router picks a model for each request. Your own keys and local models sit in the same picker.
Team mode and parallel work
The Team leader splits the job among member agents, each working in its own worktree at the same time. A member waiting for approval shows on the board.
Approve on iPhone
Command approvals arrive on your iPhone. Approve with Face ID, compare the changed files on the phone, then merge.
How it runs
- 1
Ask
Pick an assistant or just say it. Korean first across the UI, docs and replies; English in full.
- 2
Run
The agent reads and edits files and runs commands. Every turn is a checkpoint.
- 3
Approve
Risky commands ask on the desktop or your iPhone. Commands on your allowlist never ask.
- 4
Check
Read the changed files and the test result; if it is not right, go back to any turn.
Go one level deeper
One e-mail makes a free account, and your own API keys or local models work as they are.