Flow editions
Flow is one platform delivered through several editions (form factors). Every edition shares the same Rust orchestration core and catalog-driven node registry. What differs is the shell, the execution locus, and the client surface.
| Edition | Surface | Execution locus | Status |
|---|---|---|---|
| Flow Studio | Desktop app (Tauri shell) | Local machine | Current |
| Flow Code | VS Code sidebar | Local machine (in VS Code) | Shipped |
| Flow Code (JetBrains) | IntelliJ Platform plugin | Local machine (in the IDE) | Planned |
| Flow Server | Browser (remote) | Private cloud instance | Available |
| Flow CLI | Terminal / CI | Local machine, headless | Preview |
| Flow Mobile | Mobile app | - | Exploratory |
| Flow Edge | Edge deployment | - | Exploratory |
Flow Studio
Section titled “Flow Studio”Flow Studio is the desktop edition. It is the shipped foundation the other editions build on. Local inference stays zero-egress by default. Cloud AI and service-node APIs are opt-in carve-outs that are off by default. They are gated by settings, and credentials are kept in the OS keyring. Studio documentation.
Flow Code
Section titled “Flow Code”Flow Code is the VS Code edition. It is a native sidebar that generates a Flow DSL from a prompt, runs it on the shared orchestration core in the open editor, and lets the model monitor and repair the run. The workspace is pinned to the open VS Code folder, and the engine owns every file and shell action. The core runs out-of-process as a JSON-RPC sidecar. Flow Code documentation.
Flow Code (JetBrains)
Section titled “Flow Code (JetBrains)”This is the same Flow Code edition re-hosted as an IntelliJ Platform plugin, so it runs natively in IntelliJ IDEA, PyCharm, WebStorm, GoLand, and the other JetBrains IDEs. It is a parallel host rather than a fork: it reuses the same out-of-process core sidecar and the same webview bundles unchanged, re-hosts them in JCEF, and rebuilds only the editor-host glue in Kotlin. It is planned and not yet shipped. The VS Code edition remains the shipped reference.
Flow Server
Section titled “Flow Server”Flow Server is the same platform as Flow Studio. It is hosted on your own private cloud instance and reached remotely in a browser instead of running on your machine. The engine, node registry, and local-inference model are identical. Inference still runs inside the private instance and is never sent to a third-party model service, so the zero-egress posture holds. Only the compute location and the access path change. Server documentation.
Flow CLI
Section titled “Flow CLI”Flow CLI is a headless runner that executes Flow DSL directly on the shared
Rust core, without the desktop shell, for terminals and CI. It also includes a
full-screen TUI. It is distinct from the planned Flow export / Code Generator,
which transpiles a flow away from the runtime into a standalone script. Flow
CLI keeps the runtime and runs the .flow document natively. CLI
documentation.
Flow Mobile and Flow Edge
Section titled “Flow Mobile and Flow Edge”These are exploratory editions in early planning. They slot into the same structure rather than forking it. Flow Mobile is the mono UI in a Tauri mobile shell, with IPC to an in-process core, like Studio. Flow Edge is the server binary deployed at the edge, serving the mono UI over HTTP/SSE. It works like Flow Server but in a different locus.
Shared settings
Section titled “Shared settings”Every edition reads and writes the same ~/.flow-studio/settings.json and
surfaces those settings through its own interface, so a change made in one
edition is read by the others. See the
settings reference.
One UI, one backend
Section titled “One UI, one backend”Every edition is built from two shared pillars instead of per-edition forks:
- The mono UI is the single React app every edition renders. Desktop and mobile render it in Tauri, and Server and Edge render it in the browser.
- The mono backend is the canonical HTTP/SSE server over the shared application core for the server-class editions.
A host facade in the UI is the seam that lets the one UI target either transport, selected at runtime. The transport is either Tauri IPC, for desktop and mobile with an in-process core, or HTTP/SSE, for browser and edge. The managed local-model engine is fetched once per host and reused by every edition on it.
| Edition | UI surface | Transport | Backend |
|---|---|---|---|
| Studio (desktop) | mono UI in Tauri | Tauri IPC | core in-process |
| Code (VS Code) | native sidebar | JSON-RPC sidecar | core out-of-process |
| Server (browser) | mono UI in browser | HTTP/SSE | server binary |
| Edge | mono UI SPA at edge | HTTP/SSE | server binary (deployed) |
| CLI | - | direct | core in-process |
| Mobile | mono UI in Tauri mobile | Tauri IPC | core in-process |