stagewise: an agentic IDE that reads your tab's console and debugger
The Open Source Agentic IDE. Create and orchestrate coding agents, show app previews and run git workflows. Use your favorite models across all providers.
At a glance
- What is it?
- stagewise bundles a coding agent into a browser workspace so the agent can see the page you are looking at, its console and its debugger. The AGPL-3.0 licence and a bring-your-own-key model make it a real option for some teams and a poor fit for others.
- Who is it for?
- Adopt stagewise if your work is front-end debugging or reverse-engineering a page's components and you want an agent that can read the console and debugger of the tab you are on, and if you can accept AGPL-3.0 terms for how you distribute anything derived from it. Do not adopt it if your team cannot take on a copyleft obligation, or if you need a stable pinned release, since the project publishes nightly builds alongside tagged versions.
- Can I use it commercially?
- Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
- Is it still maintained?
- Yes. The repository last received commits 2 days ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What stagewise is actually for
The README describes stagewise as an open source agentic IDE for developers with a coding agent built right in. The distinguishing claim is not that it has an agent, since most editors now do. It is that the agent has full access to the tab's console and debugger. That changes what the agent can be asked to do. Instead of pasting an error into a chat window, the agent can read the console output itself while you look at the page.
The second use case the README lists is reverse-engineering: pulling apart any website's components, style systems and color palettes. That is a narrower job than general application development, and it shapes who the tool suits. If you spend your day inside a design system trying to work out how another site built a component, the combination of a live page and an agent that can inspect it is the point. If you spend your day in backend services, the browser-centric design is overhead.
The README also separates temporary test changes from connecting a codebase for permanent edits. That distinction matters more than it first appears, because it means the tool is usable against a page you do not own, at least for the temporary path, without wiring up a repository.
The mechanism: a browser workspace with an agent attached
Based on the repository layout and the README, stagewise is a TypeScript application distributed as a desktop download from stagewise.io, with an apps/website directory in the repository holding the marketing site. The README's own framing is browse and build in the same tool, which tells you the architecture is a browser surface plus an agent process, not an editor with a preview pane bolted on.
Two capabilities follow from that. The agent can read the console and debugger of the active tab, and there is an IDE integration to view and apply code changes in your editor of choice. So the edit itself may land in your existing editor rather than in stagewise. The flow is roughly: you browse, the agent observes, the agent proposes or applies changes, and you review them where you already work.
What the supplied material does not describe is how the agent is sandboxed, what it can read beyond the active tab, or how the connection between the browser surface and the agent process is secured. Those are the questions I would want answered before pointing this at a page behind a login. The README simply does not cover them.
Getting it running and pointing it at a model
Installation is not a package manager command. The README says to download stagewise from stagewise.io and follow the short onboarding guide to set up your account. There is no npm install line in the material, so I am not going to invent one.
Model access has three paths. The first is a stagewise Account, with a Free tier at $0 per month limited to three standard models the README calls Default, Quick and Smart, a Pro tier at $20 per month with access to all models, and an Ultra tier at $200 per month with 15x higher limits than Pro. The second is connecting an existing coding subscription with a single API key. The README lists Kimi via Moonshot AI, Qwen Coding Plan via Alibaba DashScope, MiniMax, Xiaomi MiMo and Mistral, each with a dashboard link for obtaining the key. The third is bring your own API key, where the README names OpenRouter as the single-key route to 345+ models and notes that custom providers are supported, including local inference, with custom model definitions.
That third path is the one worth reading twice. Registering a custom provider and defining custom models is what makes local inference possible, and local inference is often the reason a team looks at an open source agentic IDE in the first place.
The licence is the decision, not a footnote
stagewise is developed by stagewise GmbH and offered under AGPLv3. The README points to the GNU licence FAQ and offers a sales contact for use cases outside the scope permitted by AGPLv3.
AGPL-3.0 is a copyleft licence with a network clause. The practical consequence is that if you modify stagewise and let users interact with it over a network, the source of your modified version generally has to be offered to those users. For an individual developer using the tool locally, this is unlikely to bite. For a company that wants to embed the agent in an internal or customer-facing service and keep its changes private, the README's own sales contact line is the signal that this is a commercial conversation, not a free one.
I am not giving legal advice, and the exact obligations depend on what you do with the code. The point is that the licence choice is deliberate and the vendor has already anticipated the case where it does not fit. Treat the licence as a gating question you answer before the technical evaluation, not after.
Where stagewise is the wrong tool
The clearest limitation is the one the project advertises as a feature. An agent with full access to a tab's console and debugger is an agent with access to whatever that tab contains. The README does not describe a permission model, a domain allowlist, or a confirmation step before the agent reads debugger state. If your work involves authenticated sessions, customer data in the DOM, or internal tools, that gap is a real one and you should assume it is unaddressed until you find documentation saying otherwise.
A second limitation is release cadence. The recent releases include stagewise@1.29.0 alongside nightly builds such as 1.29.1-nightly20260815c001 and 1.28.1-nightly20260815c001, published within hours of each other on the same day. Nightly tags in the release list mean the project ships fast. That is good for feature velocity and bad for anyone who wants a version they can pin and forget. If your environment requires change control around developer tooling, the nightly stream is friction you will have to manage.
The third case is simply mismatched work. If your day is API design, database migrations or infrastructure, a tool built around a live page and its debugger is not where your problems are.
How it differs from wiring an agent into your existing editor
The obvious alternative is to keep your current editor and add an agent extension to it, pointing the extension at the same provider keys stagewise accepts. That approach reuses an environment you already know and avoids a second application.
The difference in approach is where the agent's context comes from. An editor-based agent sees your files, your terminal and your diagnostics. It does not, by default, see the console output of the page running in your browser, and it cannot inspect debugger state. stagewise inverts that: the browser surface is primary and the editor integration exists to view and apply changes. The README's phrasing about viewing and applying code changes in your favorite editor is consistent with stagewise not trying to replace the editor wholesale.
So the choice is not which agent is smarter. It is whether the thing you need the agent to observe lives in the file tree or in the running page. If it lives in the running page, the editor-extension route means you keep copying console output by hand, which is exactly the loop stagewise is built to remove.
Maintenance, upgrades and what to verify
Maintenance cost here is mostly the cost of staying current. The repository is active, with a last push in September 2026 and a steady release stream, and the nightly tags indicate the main branch moves continuously. A team adopting stagewise should expect to update it on a schedule rather than install once.
There is also a configuration surface to maintain. Each provider you connect has its own API key and dashboard, and the README lists separate dashboards for Moonshot AI, Alibaba DashScope, MiniMax, Xiaomi MiMo, Mistral and OpenRouter. If you register custom providers for local inference, those definitions are yours to keep working as the application changes.
The upgrade path itself is not documented in the supplied material. There is no migration guide, no changelog excerpt and no statement about backwards compatibility between 1.28 and 1.29. That is a gap. Before adopting, read the release notes for the versions you would install and check whether the AGPL-3.0 LICENSE file in the repository matches what the README states, since the README is the only licence source given here.
Editorial conclusion
Adopt stagewise if your work is front-end debugging or reverse-engineering a page's components and you want an agent that can read the console and debugger of the tab you are on, and if you can accept AGPL-3.0 terms for how you distribute anything derived from it. Do not adopt it if your team cannot take on a copyleft obligation, or if you need a stable pinned release, since the project publishes nightly builds alongside tagged versions. Before committing, confirm the current licence text in the repository's LICENSE file, the price of the tier you actually need, and whether your editor is among those the IDE integration supports.
Community notes