Multiple Agents, One Editor
Run agents in parallel, manage their workloads, and review results without interrupting your flow.
An agentic workflow with the power of an editor
Keep the mechanics simple: start threads, choose the right agent, decide how isolated the work should be, and review results without leaving the editor.
Start a thread for each line of work
Refactor in one thread, generate tests in another, and investigate bugs in a third.
Choose the right agent for each thread
Mix and match agents in the same workspace, based on the work you need done.
Decide when to isolate work
Some tasks can share your current working copy. Others should run in their own worktree. Zed lets you make that choice per thread.
Review, compare, and merge back into your flow
As threads finish, review edits, compare approaches, and keep moving without context switching away from the editor.
Use your favorite model
Pick the setup that fits how you work. Zed connects without getting in the way.
Learn about Zed ProHosted by Zed
Use the best models, like Claude Opus and GPT-5.4, with simple pricing and no API keys to manage.
Run locally
Connect models through Ollama and keep everything on your hardware. Nothing leaves your machine.
Bring your own keys
Plug in API keys from any provider. Use your existing accounts, your existing budgets.
Continue agent conversations in Zed
Bring any ACP-powered agent thread into Zed and continue where you left off.
Learn about ACPImport external agent threads
Add any historical ACP-compatible agent thread into your Zed thread history.
Explore a growing list of agents
Experiment with different agents by exploring the full list available in the ACP registry.









