What is Ensemble?
Ensemble is a cloud workspace for shipping changes to an existing codebase. Product managers, designers and founders describe a change in the browser and get a live preview environment: a full running copy of the app with seeded data, at its own URL. Coding agents like Claude Code and Cursor connect to Ensemble over MCP for codebase and product context.
What Ensemble does
Ensemble does two things. It gives every change a live, shareable environment, and it gives every AI working on the code an understanding of the product.
Those two halves solve one problem. More people are writing code than ever: developers with AI agents, and PMs, designers and founders who now make changes themselves. But two things are still stuck on individual laptops. The running app lives on localhost, so nobody else can see a change. And the understanding of the codebase lives in engineers' heads, so AI agents guess. Ensemble moves both into a shared place.
It has three parts:
- Ensemble Studio is the browser workspace: chat, a hot-reloading preview and logs, with nothing installed locally. It's where PMs, designers and founders make changes.
- Ensemble Previews give every branch or ticket a full running stack (your services, Postgres, Redis and seeded data) at its own URL. Previews sleep when idle and sit behind a login by default.
- Ensemble Context is a codebase graph plus product knowledge, re-indexed when protected branches change. People and coding agents query it instead of re-reading the repository.
There are three ways in: the web workspace, the Linear integration, and the MCP server for Claude Code, Cursor and similar tools. A developer who only ever uses Ensemble through MCP gets the full value of Context without opening the web app.
Who Ensemble is for
Ensemble is built first for product people and founders who want to ship changes themselves, and second for the engineers who work alongside them.
| Who | What they want | What Ensemble gives them |
|---|---|---|
| Product manager | Ship a change, or hand over one that can't be misread | Studio, and Linear tickets that become previews |
| Founder | Change the product without waiting on engineering | Studio and previews, with no local setup |
| Designer | Change UI and see it in the real app | A live preview with real data |
| Developer | Better output from their coding agent | Codebase and product context over MCP |
| Reviewer | Try a change before it merges | A link to a running, seeded environment |
| Eng lead | Let non-engineers ship without risk | Isolated previews and secrets, private links, review before protected branches |
How Ensemble works
You connect a repository, declare the environment once, and every branch after that gets its own running preview.
- Connect your GitHub repos. One repo or many, any stack.
- Declare the environment. The services a preview runs, the datastores they need, and the seed data to load.
- Ensemble indexes the code. It builds a graph of the codebase and generates product knowledge from it, and re-indexes when protected branches change.
- Make a change. Describe it in Studio, or start from a Linear ticket. An AI agent edits a branch, and the change runs on a preview at its own URL.
- Share and review. Send the link to reviewers, who don't need a paid seat. An engineer reviews the pull request before anything reaches a protected branch.
Developers can skip straight to step 3: add the Ensemble MCP server to Claude Code or Cursor, and their agent can ask questions about the codebase instead of crawling it.
How Ensemble differs from other tools
Preview hosts give a URL to people who already run the code, and app generators start a new project. Ensemble lets people with no local setup change the app that already exists.
- Preview hosts such as Vercel, Railway and Render give each pull request a preview URL, and they're excellent at it. Ensemble adds a workspace for non-engineers and codebase context for agents. See Vercel preview deployments vs full-stack previews and Railway vs Render preview environments.
- App generators such as Lovable, v0 and Bolt are the fastest way to start something new. Ensemble is for changing a product that already exists. See Lovable alternative for an existing codebase.
- Coding agents such as Claude Code and Cursor are complements. They write the code; Ensemble tells them what they can't see in the files. See why coding agents re-read the repo.
What Ensemble costs
Ensemble starts at $0 on Pay as you go, and Team is $30 per builder seat per month. Reviewers are free.
- Pay as you go: $0 a month, no card needed. 100 credits included (about 40,000 lines), one preview awake at a time, then prepaid top-ups at $20 per 100 credits. Hibernated previews are free.
- Team: $30 per builder seat per month, 3-seat minimum. 500 credits per seat (about 200,000 lines), pooled, 2 previews awake per seat and custom preview domains. Overage is $2.50 per 10,000 lines.
- Pro: $50 per builder seat per month, 5-seat minimum. 1,000 credits per seat (about 400,000 lines), 5 previews awake per seat, larger machines and SSO access. Overage is $0.08 per credit.
A builder is anyone who chats, edits or creates previews. Everyone who only opens links is free. See what is a builder seat and the pricing page.
FAQ
Is Ensemble an AI coding agent?
No. Ensemble gives your coding agent context and gives everyone a running preview of its work. You keep Claude Code, Cursor or whichever agent you use today.
Does Ensemble host my production app?
No. Ensemble previews are for seeing and reviewing changes before they merge. Production stays on the host you use today.
Does Ensemble work with my existing codebase?
Yes. Ensemble is built for existing codebases. You connect your GitHub repos and declare the services the app needs.
Do I need to install anything?
Not to use Studio: it runs in the browser. Developers who want context in their editor add the Ensemble MCP server to Claude Code or Cursor.
Is Ensemble related to ensemble learning in machine learning?
No. Ensemble is a developer product, a cloud workspace for shipping code changes. The name is unrelated to ensemble methods in machine learning.