Cezar CLI
A CLI opens the Cezar dashboard for coding-agent tasks, with separate sessions, Git worktrees and subtask reports
- 949 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Adapters can execute commands and modify files with broad permissions
- Some agents receive permissions automatically without a human prompt
- Worktrees and connected tools can receive repository secrets
Install
- Works in
Manual install
npx cezar-runRun in the project directory. Actual work needs an installed, authenticated agent.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Adapters can execute commands and modify files with broad permissions
- Some agents receive permissions automatically without a human prompt
- Worktrees and connected tools can receive repository secrets
This is third-party code. Review the repository files before installing.
What it does
Cezar opens a local dashboard for several coding agents. Create a task and choose an installed client such as Claude Code or Codex. The dashboard shows session progress and lets you send clarification to the executor. Separate Git worktrees support parallel work. A subtask gets its own session and report. Its parent supplies the goal and boundaries so the executor can work without inheriting the full main-session history. A review can target a particular branch or run. Reports connect findings to changes but do not merge code by themselves. Short tasks can run without a browser; subtask dispatch in that mode depends on a running server.
Who it is for. Agent-task coordinators separating working branches and reviewing parallel results before merging.
Good fit when
- You want one dashboard for different installed agent CLIs
- Parallel tasks need separate branches and worktrees
- You need to delegate a bounded task and receive a report
- A particular branch needs review by another executor
Not a fit when
- You only need a short terminal session
- Executors need isolation but you cannot review each adapter’s permissions
- You need every agent command to require confirmation
Example request
Create one Cezar test task for my repository. First show adapter permissions and the worktree plan, then propose a separate branch review.Limitations
Node.js 20 or later and an installed agent with model access are required; worktrees need Git and GitHub operations need GitHub CLI. Codex defaults to danger-full-access and approvalPolicy: never, Copilot gets --allow-all-tools, and Junie/OpenCode approve permission requests automatically. A worktree does not isolate the process or restrict access outside the project. Cursor token and cost reports are incomplete; a dashboard zero can mean missing data.
How to disable. Stop active tasks and the Cezar server. Disable any separately installed system service too. Preserve needed history and inspect Git worktrees before removal. Repository changes and GitHub publications need separate reversal.
FAQ
How do I open Cezar for a project?
Run `npx cezar-run` in the project directory. The dashboard opens at `http://localhost:4321` or the next free port.
Does it need a global installation?
No. The main route uses npx. The older `npx cezar-cli` name starts the same package.
How can I restrict networking for the Codex adapter?
CEZ_CODEX_NETWORK=0 selects workspace-write and disables networking for the Codex adapter. It does not change other agents’ permissions.
Does a subtask report merge its changes?
No. Status and reports describe the result. Review them against the branch and evidence before deciding on a merge.
Why can cost display as zero?
Some adapters do not return complete usage information, particularly Cursor’s non-interactive print mode. That zero does not establish free execution.
What license and separate costs apply?
Code uses MIT. Agent subscriptions and server resources have separate costs.
Discussion
No questions or reviews yet.
Related
CLI toolsA local control panel for coding agents: it runs Claude Code and Codex, keeps a task queue and merges verified work into main.
A workspace for parallel Claude Code and Codex tasks, with separate Git worktrees, terminals and change review
A memory and coordination layer for coding agents: specs and issues live in git while agents work in parallel
A local dashboard for many Claude Code sessions: each job gets its own context and todo, agents spawn child agents and message each other
A planner for Claude Code, Codex and OpenCode: one goal becomes a task plan and independent tasks run in parallel
DevChain runs a team of AI coding agents on your own machine with a shared board, chat, code review and live terminals.
README in short
Cezar connects existing agent CLIs through separate adapters. It supports a local dashboard and headless operation. Subtasks share a parent budget and return work reports.