Run Cursor in ace
Cursor's coding agent, run from its command-line tool. ace runs it through your own signed-in CLI and adds what a terminal can't: every task in its own thread, a status that covers its subagents, and tools to see and test what it builds.
What Cursor gets in ace
- Agent orchestrationDelegate from any agent to any provider
- Computer useAgents read apps through accessibility, then click and type
- FleetAgents start threads and open browsers on linked machines
- iPhone and iPadLive Activities that end only when the whole agent tree is done
- Remote accessSign in with email, Google or GitHub
- Accounts and limitsEach account signs in through the CLI's own login, kept apart
- Browser useOne shared browser: the same tabs and logins for you and agents
- DevicesLive screen, UI tree, logs and recordings
- Worktrees and reviewOne git worktree per thread, created for you
- Thread statusWorking, Approve, Waiting, Failed and Done
- Run Cursor's agent beside Claude Code and Codex, each thread in its own worktree.
Set up Cursor
- Install Cursor and sign in, the way its own instructions say. ace uses that login as it is.
- Download ace for your computer and open a project folder.
- Start a thread and choose Cursor. Give it its own worktree if other agents share the repository.
Questions about Cursor in ace
Does ace need a Cursor API key?
No. ace starts Cursor through its own command-line tool, signed in the way you already use it, so your existing plan keeps working. ace never asks for, stores or proxies credentials.
Can I run Cursor beside other agents?
Yes. Run Cursor, Claude Code, Codex, Cursor and others side by side, each thread in its own git worktree, and hand a thread from one agent to another with its context.
Can Cursor hand work to other agents?
Yes. Cursor can delegate tasks to agents on other providers and on your other machines with ace's delegate tool, and wait for their results.
Can Cursor use a browser and my desktop apps in ace?
Yes. ace gives Cursor the same MCP tools as every other agent: a real browser, computer use, phone simulators, and delegating to agents on other machines.
Other agents
Claude Code
Anthropic's coding agent, run from its own command-line tool.
Codex
OpenAI's coding agent, run from its own command-line tool.
OpenCode
The open-source coding agent that works with many model providers.
Antigravity
Google's Antigravity coding agent.
Pi
The Pi coding agent, run from its own command-line tool.
Free, open source, and yours to run.
Install ace on your computer, point it at a project, and start a thread with any agent you have logged in.
v0.0.19 for macOS on Apple silicon, 239 MB. v0.0.19 for Windows on x64, 208 MB. v0.0.19 for Linux on x64 (AppImage). Release notes
Star on GitHub