Human and AI agent collaboration

Work with AI agents like teammates.

Ensemble is a workspace for collaborating with AI agents and LLMs. You and your agents share one board, one set of notes and one context. Agents pick up the work you assign, ask when they need you, and wait for your approval before anything leaves.

One board for you and your agents

Lanes for You, Agent and Needs you. See what the agent is doing, how far it is and what it plans next. Give a task to the agent, keep it yourself, or give it to a person in a shared space.

Every agent asks in one place

Claude Code, Codex, GitHub Copilot and Cursor send their permission prompts and questions to Needs me, with their own options and room for your own answer. Answer once, for the session, or always, from the web, your phone or the terminal. The hook never allows on its own.

Agents run on your computer, sandboxed

The ensemble CLI runs assigned tasks in a folder you choose, inside the computer's own sandbox: Seatbelt on macOS, bubblewrap or Landlock on Linux, a separate sandbox account on Windows. No Docker. Hand the task to Ensemble's agent or to Claude Code with your own sign-in.

Approvals first

Anything that leaves your workspace waits for a yes from you. Pause the agent, reject a draft, or undo its writes. In a folder you trust, let it run unattended and read every decision it made at the end.

A record of everything an agent did

Every agent action lands in a tamper-evident, hash-chained ledger with its inputs and approvals, so a run can be replayed and checked later.

Any model, your key

Run agents on Gemini, OpenAI, Claude, OpenRouter, Ollama or GitHub Copilot with your own key. Keys are encrypted and never leave the model runtime.

FAQ

Questions, answered.

What does it mean to work with AI agents?

Working with AI agents means handing real tasks to software that can plan and act, not just answer a chat. In Ensemble you assign a task on a shared board, the agent works with the same notes and context you have, and it comes back to you for decisions and approvals.

Which AI agents work with Ensemble?

Ensemble has its own agent that runs on Gemini, OpenAI, Claude, OpenRouter, Ollama or GitHub Copilot with your key, and it can hand coding tasks to Claude Code with your own Claude sign-in. Permission prompts from Claude Code, Codex, GitHub Copilot and Cursor come to Needs me, and editors read your context over MCP.

Can an AI agent act without my approval?

Not for anything that leaves your workspace: that waits for your approval. Commands run in a sandbox inside a folder you chose, and only a folder you trust can run unattended; paths outside it are refused.

Can people and AI agents collaborate in the same workspace?

Yes. Share a space or a single page, task, board or diagram with people. Everyone sees the agent's runs and who is on each page, and a member's agent tasks run on their own computer.

Keep reading

More of the workspace.

Bring your agent into the room.

Create a free account, save your notes and context, and bring your own model key when you are ready to use an agent.