StructSmith

Model, document and share software architecture — locally, and with your AI client.

_{▶ 90-second demo — model a payments system, name the
relationships as you draw them, group containers in a boundary, then drill into one of them.}

A local-first, open-source tool for modelling software architecture — a self-hosted alternative to Structurizr with a modern React Flow editor and **MCP built in**, so your AI client works on exactly the same model you do. > The diagram is **not** the source of truth. The semantic architecture model is. > React Flow only visualises and lays out that model. Useful for software architecture, the C4 model, presales, solution architecture, architecture workshops, documenting systems, designing integrations — and for working on architecture together with an AI assistant. - MIT licensed, no account, no cloud, no telemetry — it works fully offline - One container, one SQLite file - REST API, MCP and the UI all sit on the same domain layer --- ## Quick start Stable releases and automatically generated release notes are available under [GitHub Releases](https://github.com/dziksu/StructSmith/releases). After the first automated release, `latest` tracks the latest published release; use `:vX.Y.Z` instead of `:latest` to pin a specific published version. ### Run the public image With Docker installed, run: ```bash docker run -d --name structsmith \ -p 127.0.0.1:8090:8080 \ -v structsmith-data:/data \ --restart unless-stopped \ ghcr.io/dziksu/structsmith:latest ``` Then open . No repository clone or local build is needed. Your workspaces persist in the `structsmith-data` Docker volume. Choose the launcher below if you want the built-in chat to use CLIs installed on your computer. Plain `docker run` and Compose run the server inside the container, which cannot execute the host's programs or use its CLI login automatically. ### Docker with host-installed agents, without cloning On macOS or Linux with glibc (arm64/x64), install Docker, curl and at least one signed-in Codex, Claude Code or Copilot CLI. The local launcher downloads the matching Docker release and runs chat on your computer: ```bash curl -fsSL https://github.com/dziksu/StructSmith/releases/latest/download/structsmith-local-install.sh | sh ``` The release installer verifies the binary's SHA-256 checksum. No Git clone, Node.js or Bun installation is required. The installer is attached after the matching image is published. If this URL returns 404, check the release's image and local launcher publication jobs; release notes alone do not mean the files are ready to download. Open and choose **Agent settings** to select your installed CLI, model and reasoning effort. The launcher uses port 8090 for UI/chat and 8091 for the private Docker backend, both on loopback. It reuses the `structsmith-data` volume, keeps chat under `~/.local/share/structsmith/chat`, and leaves CLI credentials on the host. An optional source directory is a host path. Stop an existing instance using those ports or the same data volume before starting this mode. For a Compose deployment, pass its actual volume name with `--volume NAME` if you want to reuse the existing model. The browser talks to the localhost helper. It serves chat directly and proxies the editor to Docker; agent changes still pass through the domain's MCP tools, revision checks and snapshots. No agent credentials or host executable need to be mounted into the container. Keep its terminal open. Ctrl+C stops the helper, active agents and managed container; it retains the model volume and chat history. Subsequent starts use: ```bash ~/.local/share/structsmith/structsmith-local ``` For a different port, pass `--port 8095` (the backend uses 8096). In another terminal, `~/.local/share/structsmith/structsmith-local status` or `stop` controls the saved profile. Use `--volume NAME` to select another Docker model volume, `--read-only` to disable architecture changes, or `--help` for all options. The launcher refuses to replace containers that belong to another deployment. It also refuses a volume used by another running container. To upgrade, stop the launcher and rerun the installer command; the downloaded helper and image share the same release version, while the existing model and conversations are retained. For testing changes from a checkout, build the matching local image and start it: ```bash docker build -t structsmith-local:dev . bun run docker:local --image structsmith-local:dev ``` `bun run docker:local` by itself uses the version pinned release image. The helper and image versions must match. More details are in the [local chat guide](docs/LOCAL_AGENT_CHAT.md). ### Build from source with Docker Compose ```bash git clone https://github.com/dziksu/StructSmith.git structsmith cd structsmith docker compose up -d --build ``` Then open . ### Local development ```bash git clone https://github.com/dziksu/StructSmith.git structsmith cd structsmith bun install --frozen-lockfile bun run dev ``` `bun run dev` starts both processes: the Bun/Express API on `:3000` and the Vite dev server on , which proxies `/api` and `/mcp` to the backend. The SQLite database is created and migrated automatically — no Docker, Postgres, Redis, Java or Python required for development. The static project website lives in `apps/site`. Run it separately on with: ```bash bun run dev:site ``` `bun run build:site` creates the GitHub Pages artifact in `apps/site/dist`. Changes to the site, shared public assets, or its workflow are published automatically from `main` by the `GitHub Pages` action. Before the first deployment, select **GitHub Actions** as the source in the repository's **Settings → Pages**; later pushes need no manual publishing step. The build prerenders the complete page with React's server renderer and inlines its CSS into `index.html`; the published page has no client-side React runtime or JavaScript bundle. --- ## Endpoints | URL | Purpose | | --- | --- | | `http://localhost:8090/` | Web UI | | `http://localhost:8090/api/...` | REST API | | `http://localhost:8090/mcp` | MCP (Streamable HTTP) | | `http://localhost:8090/health` | Health check | In development the same endpoints live on `http://localhost:3000` (and `:5173` for the UI). --- ## Connecting an AI client (MCP) MCP is a core feature, not an add-on: it starts with the backend and is served from the same process. The transport is **Streamable HTTP** (the deprecated SSE transport is not implemented). ### Quick install by client Start StructSmith first, then use the command for your client. The examples assume the public Docker image is available at `http://localhost:8090`. **Codex CLI, Codex Desktop and the Codex IDE extension** share the same MCP configuration: ```bash codex mcp add structsmith --url http://localhost:8090/mcp ``` In Codex Desktop you can instead open **Settings → MCP servers → Add server**, choose **Streamable HTTP**, and paste the endpoint. Restart the current session after changing MCP configuration. **Claude Code:** ```bash claude mcp add --transport http structsmith http://localhost:8090/mcp ``` **GitHub Copilot CLI:** ```bash copilot mcp add --transport http structsmith http://localhost:8090/mcp ``` **VS Code with GitHub Copilot** — add `.vscode/mcp.json` to the project where you want to use StructSmith: ```json { "servers": { "structsmith": { "type": "http", "url": "http://localhost:8090/mcp" } } } ``` Then run **MCP: List Servers** and start `structsmith`; use Copilot Chat in Agent mode. **Claude Desktop** needs a local STDIO bridge to reach a StructSmith instance on `localhost`. See the [AI client installation guide](docs/AI_CLIENTS.md#claude-desktop) for a copy-ready configuration. A public HTTPS deployment can instead be added under **Customize → Connectors → Add custom connector**, but Claude connects to it from Anthropic's cloud — `localhost` will not work there. For Cursor, Windsurf, JetBrains IDEs and other MCP clients, point a native Streamable HTTP connection at the endpoint or use this common configuration shape: ```json { "mcpServers": { "structsmith": { "type": "http", "url": "http://localhost:8090/mcp" } } } ``` For full client-specific instructions, token authentication, local-development ports and troubleshooting, see **[Install StructSmith in AI clients](docs/AI_CLIENTS.md)**. The MCP page in StructSmith also generates copy-ready snippets using the live URL, including custom ports. ### Manual Codex configuration Codex can also load StructSmith automatically from a trusted project's `.codex/config.toml`: ```toml [mcp_servers.structsmith] url = "http://localhost:8090/mcp" default_tools_approval_mode = "writes" tool_timeout_sec = 120 ``` For clients that only speak stdio: ```bash bun run mcp:stdio ``` The UI shows the live endpoint, the tool list and the read-only state under **MCP** in the top bar. ### What the AI can do Tools cover workspaces, the model, elements, relationships, views, view-owned boundaries, presentation settings, saved layout, presales records, snapshots and export. The preferred way to make a larger change is a single call to `model_apply_operations`, which is atomic, revision-guarded and takes an automatic snapshot first: - Start with `workspace_list`, then `workspace_inspect` for one complete, validated context packet. View membership, boundaries and settings are always included; request `includeLayouts` when coordinates or relationship bend points are needed. - Call `modeling_guide` for allowed values, C4 rules, relationship lifting and the recommended workflow instead of reading StructSmith source code. - Run `model_preview_operations` before a large batch. It uses the real operation engine and validator, then rolls the transaction back. - Apply the accepted batch with `model_apply_operations` and finish with `model_validate`. - References copied from the UI are one-line `StructSmithRef` JSON payloads. Pass their `workspaceId`, `type` and `targetId` to `reference_resolve` to fetch the exact object and context. ```jsonc { "workspaceId": "example-client-portal", "label": "Add asynchronous invoice processing", "operations": [ { "op": "createElement", "ref": "queue", "data": { "kind": "container", "role": "queue", "name": "Invoice Queue", "technology": "SQS" } }, { "op": "createElement", "ref": "worker", "data": { "kind": "container", "role": "worker", "name": "Invoice Worker" } }, { "op": "createRelationship", "data": { "sourceElementId": "backend-api", "targetElementId": "@queue", "description": "Publishes invoice jobs to", "interactionStyle": "async" } }, { "op": "createRelationship", "data": { "sourceElementId": "@queue", "targetElementId": "@worker", "description": "Delivers jobs to", "interactionStyle": "async" } }, { "op": "setViewElements", "viewId": "containers-view-id", "elementIds": ["@queue", "@worker"] }, { "op": "autoLayoutView", "viewId": "containers-view-id", "direction": "LR", "algorithm": "dagre" } ] } ``` `ref` gives a new entity a local alias; later operations reference it as `@alias`. The whole batch runs in one SQLite transaction — it either lands completely or not at all. Boundaries are owned by a view and never become model elements or relationship endpoints. The same element can belong to different boundaries in different views or layers. An element with no boundary in the active layer remains an ordinary item in that view. Resources expose the model in an AI-friendly shape (no React internals, no CSS, no viewport data): ``` architecture://guide architecture://workspaces architecture://workspace/{workspaceId} architecture://workspace/{workspaceId}/inspection architecture://workspace/{workspaceId}/model architecture://workspace/{workspaceId}/views architecture://workspace/{workspaceId}/view/{viewId} architecture://workspace/{workspaceId}/records ``` Prompt templates: `review_architecture`, `create_presales_architecture`, `identify_architecture_risks`, `identify_unknowns`, `review_security`, `review_scalability`. When an AI changes the model, the browser is notified over SSE (`GET /api/events`) and the diagram updates on its own — no refresh needed. --- ## The model ``` Workspace ├── Elements person · softwareSystem · container · component · │ deploymentNode · infrastructureNode · custom ├── Relationships sync · async · event · data · dependency · custom ├── Views landscape · systemContext · container · component · │ │ deployment · custom │ └── Boundaries deployment · security · compliance · ownership · custom ├── Records assumption · risk · unknown · requirement · decision · note └── Snapshots ``` StructSmith models itself — this diagram is the container view of the example self-model, exported straight from the tool with `export_mermaid`: ```mermaid flowchart LR architect(["Architect"]) ai_client["AI Client
Claude, Cursor, any MCP client"] subgraph structsmith["StructSmith"] direction LR web_ui["Web UI
React, Vite, React Flow"] rest_api["REST API
Bun, Express"] mcp_server["MCP Server
MCP SDK, Streamable HTTP"] domain["Domain Services
no Express, no SQL"] event_bus["Event Bus
in-process, SSE"] repositories["Repositories
Drizzle ORM"] sqlite["SQLite
WAL, bun:sqlite"] end architect -->|"Models architecture in"| web_ui ai_client -->|"Reads and changes the model through"| mcp_server web_ui -->|"Calls"| rest_api rest_api -->|"Delegates to"| domain mcp_server -->|"Delegates to the same layer as REST"| domain domain -->|"Reads and writes through"| repositories repositories -->|"Persists to"| sqlite domain -.->|"Publishes model changes to"| event_bus event_bus -.->|"Notifies the SSE stream in"| rest_api rest_api -.->|"Pushes workspace.updated to"| web_ui ``` Three rules the implementation is built around: 1. **Positions belong to views, not to elements.** One element can appear on five diagrams without five copies of the data; moving a node never changes the architecture. 2. **Roles are not C4 levels.** `kind = container`, `role = database`, `technology = PostgreSQL` — roles drive icons, styling and filtering only. 3. **Views show implied relationships.** If a view shows a system but not its containers, traffic to those containers is lifted to the system, the way Structurizr does it. Every workspace has a `revision`. Mutating REST and MCP calls may pass `expectedRevision`; a mismatch returns `409 Conflict`, so a user and an AI editing at the same time cannot silently overwrite each other. --- ## REST API ``` GET /api/workspaces POST /api/workspaces GET /api/workspaces/:id PATCH /api/workspaces/:id DELETE /api/workspaces/:id POST /api/workspaces/import POST /api/workspaces/import/mermaid GET /api/workspaces/:id/model GET /api/workspaces/:id/document GET /api/workspaces/:id/validate GET /api/workspaces/:id/activity GET /api/workspaces/:id/export/mermaid POST /api/workspaces/:id/commands # atomic batch of operations POST /api/workspaces/:id/elements PATCH/DELETE /api/elements/:id GET/POST /api/views/:id/boundaries PATCH/DELETE /api/boundaries/:id POST /api/workspaces/:id/relationships PATCH/DELETE /api/relationships/:id GET/POST /api/workspaces/:id/views GET/PATCH/DELETE /api/views/:id PATCH /api/views/:id/layout # batched, debounced layout write POST /api/views/:id/elements POST /api/views/:id/auto-layout GET/POST /api/workspaces/:id/records PATCH/DELETE /api/records/:id GET/POST /api/workspaces/:id/snapshots POST /api/snapshots/:id/restore GET /api/events # SSE: workspace.updated GET /api/mcp-info GET /api/presets GET /api/settings ``` Errors always use the same envelope: ```json { "error": { "code": "WORKSPACE_NOT_FOUND", "message": "...", "details": {} } } ``` --- ## Editor - Desktop-first three-pane layout: explorer, canvas, inspector — all resizable - Custom React Flow nodes with icon, name, technology and a kind/role badge; external elements are visually distinct - Import existing Mermaid diagrams in `flowchart`/`graph` syntax from `.mmd` / `.mermaid` files as a semantic workspace baseline - Semantic, nested boundaries owned by a view, so the same model element can be grouped differently on deployment, security, compliance and ownership diagrams - Explorer separates the active view tree from the reusable element library; elements can be added or dragged into a view or one of its boundaries - Manual layout (debounced batch save) and automatic dagre, force, radial and grid layouts; compound dagre layout keeps boundary members together - Views: the same element on many diagrams, each with its own structure, boundaries and layout - Presales records linked to elements, with a subtle risk indicator on the canvas - Deterministic validator with error / warning / info levels - Export: semantic JSON, Mermaid, PNG, SVG. Import: native JSON and Mermaid flowcharts - Snapshots with restore; undo/redo (`⌘Z` / `⌘⇧Z`) rides on them - Command palette (`⌘K`), `F` to fit the view, `Delete`, `Escape` - Light / dark / system themes, English and Polish UI - Built-in chat with local Codex, Claude Code and GitHub Copilot CLIs: project topics, general discussions, contextual node/relationship actions, scoped architecture edits, model and Codex thinking settings, streamed replies with collapsible reasoning, and topic renaming, archiving, restoration and drag-and-drop ordering ### Chat with local agents Run StructSmith from source on the same machine as your signed-in CLI (`bun run dev`). Open **Agent chat** using the bubble in the lower right, select a project or **General · no project**, and create a topic. A project here is a StructSmith workspace; each topic shows its project and keeps that assignment when you navigate elsewhere. Right-click a node or relationship and choose **Ask agent about this** to attach it as a context chip; the inspector, view and record reference controls also offer a chat button. Review the prompt, then press **Send** or Enter. Shift+Enter adds a new line. Choose **Codex**, **Claude Code** or **GitHub Copilot** inside the topic. **Ask** is the default. **Edit architecture** enables MCP changes in that topic's project, with revision checks, activity and snapshots; the diagram updates through the existing event stream. General topics cannot edit projects. Switching agents keeps the topic's conversation and project. Open the chat gear to configure **Agent settings**: | Setting | Behavior | | --- | --- | | Default agent | Provider used for new topics; existing topics keep their selected agent | | CLI executable | Command or absolute path for each provider, such as `codex`, `claude` or `copilot` | | Model | Codex offers a list from the configured CLI plus a custom model ID; other providers accept an optional model ID. Use the CLI default to keep its configured model | | Codex reasoning effort | Thinking level for Codex; **Default** keeps the CLI's configured behavior | Model and reasoning settings apply to subsequent turns across that provider's topics. Refresh the Codex model list after changing its executable, updating the CLI or switching CLI accounts. The list comes from `codex app-server` and may differ from the desktop app's models; account access is still checked when a turn runs. Available Codex thinking levels follow the selected model's CLI catalog. For a custom model outside the catalog, compatibility depends on the installed CLI and model; changing to a known model that cannot use the selected level resets it to **Default**. The topic gear configures its title and optional absolute source directory on the server for reading local code. Manage topics from the sidebar: - **Rename:** open the topic's **…** menu and choose **Rename**. - **Archive / restore:** choose **Archive** to keep the history and project in the **Archive** tab. Use **Restore topic** there to continue the conversation. Renaming and archiving are disabled while that topic's agent is running. - **Reorder:** hold the topic body for about 350 ms, outside the **…** action button, then drag it with a mouse or touch. With keyboard focus on the topic, press Space, use the up/down arrows and press Space to drop; Escape cancels. The saved order survives reloads and server restarts, including when topics receive new messages. Replies stream into the conversation as the CLI produces them. Readable reasoning exposed by the CLI appears in a collapsed **Reasoning** panel; Codex provides reasoning summaries. Expand it to follow updates live. Both the reply and available reasoning are saved with the topic. Scrolling up keeps your reading position; returning near the bottom resumes following the response. Closing the chat or switching topics keeps the agent running, and reopening retrieves its current response. **Stop** interrupts the turn while preserving partial output and any architecture changes already applied. Agent authentication stays with the installed CLI; sign in in your terminal first. For native development, history, settings, topic order and archive status are saved under `data/agent-chat` (configurable with `AGENT_CHAT_DIR`). The local Docker launcher saves them under `~/.local/share/structsmith/chat`. Set `AGENT_CHAT_ENABLED=false` on a native backend to disable its CLI chat bridge. This feature requires a localhost connection and a recent CLI version. Use the local launcher above for a Docker deployment with host-installed agents. A plain `docker run` or Compose deployment sees only programs installed inside its container. See the [local chat guide](docs/LOCAL_AGENT_CHAT.md) for testing and implementation details. ### Mermaid import On the home page, choose **Import workspace**, drop or select a JSON backup, `.mmd`, `.mermaid`, or text file, then edit the suggested workspace name and choose **Import and open**. File selection alone does not create a workspace. The dialog uses `react-dropzone` without a native extension filter so macOS can select Mermaid files; content validation happens during import. Alternatively, call `POST /api/workspaces/import/mermaid` / MCP `import_mermaid` with `{ "source": "flowchart LR\nA[Client] -->|Uses| B[API]" }`. An optional `name` overrides the Mermaid frontmatter title. Import creates a new workspace by default. `mode: "overwrite"` requires `workspaceId` and replaces that workspace's model, views, and records; it does not merge or synchronize. MCP read-only mode does not expose this operation. Supported syntax includes `flowchart` and `graph`, all five directions, standalone and inline node declarations, traditional shapes, `@{ shape: ..., label: ... }`, Unicode and quoted labels, multiline/Markdown strings, common HTML/Mermaid entities, semicolon-separated statements, labeled and chained edges, `&` endpoint groups, dotted/thick/long links, circle/cross arrowheads, bidirectional links, edge IDs, and nested subgraphs (including links to subgraphs). A single fenced Mermaid block and YAML frontmatter with a scalar `title` are accepted. Configuration, styles, classes, click actions, accessibility directives, and invisible layout links are ignored; callbacks and external image/icon resources are never executed or loaded. Nodes and subgraphs become reusable semantic elements of kind `custom`; subgraph membership becomes parent/child containment. This is an import convention, not an inference of C4 meaning from visual shapes. Kind, role, and technology explicitly embedded in StructSmith export labels are recovered. Edge labels become descriptions; a trailing `[technology]` is recognized. Dotted links map to `async`, directed links to `sync`, and undirected links to `custom`; bidirectional links create two directed relationships. Mermaid IDs are retained as provenance in `properties["mermaid.id"]`. Layout is computed by the existing domain layout helper and stored only on the view. RL/BT imports are mirrored initially; later automatic layout uses the saved LR/TB axis. This is a semantic starting point, not a pixel-perfect or lossless round trip. Per-subgraph directions, styling, animation, exact shapes, and Markdown formatting are not reproduced. Exported view boundaries cannot reliably be distinguished from semantic subgraphs. Mermaid does not carry all StructSmith model metadata, and older exports may use parent endpoint IDs distinct from their subgraph IDs. Sequence, class, ER, state, C4, and other diagram families require separate semantic adapters and are rejected with a clear error. Unrecognized structural syntax also fails before any workspace is changed. Limits are 1,000,000 source characters, 2,000 nodes, and 10,000 relationships. The syntax scope follows the official [Mermaid flowchart documentation](https://mermaid.js.org/syntax/flowchart.html). The transport/domain integration is small; most implementation and maintenance complexity lies in syntax coverage and mapping Mermaid constructs into a semantic model. Native JSON remains the format for complete backups and lossless round trips. ### Adding a language The UI is fully translated through i18next. Drop a JSON file next to `apps/web/src/i18n/locales/en.json`, register it in `apps/web/src/i18n/index.ts` and add it to `supportedLanguages`. No copy is hard-coded in components. --- ## Configuration | Variable | Default | Meaning | | --- | --- | --- | | `PORT` | `3000` (`8080` in Docker) | HTTP port | | `HOST` | `0.0.0.0` | Bind address | | `DATABASE_PATH` | `./data/architecture.db` | SQLite file | | `MIGRATIONS_DIR` | `./migrations` | Migration folder | | `AUTH_MODE` | `none` | `none` or `token` | | `APP_TOKEN` | – | Bearer token required when `AUTH_MODE=token` | | `MCP_READ_ONLY` | `false` | When `true`, MCP exposes no mutating tools | | `SEED_EXAMPLE` | `true` | Seed the example workspace on first boot | | `APP_NAME` | `StructSmith` | Product name shown in the UI | | `AGENT_CHAT_ENABLED` | `true` | Enable the localhost-only CLI chat bridge | | `AGENT_CHAT_DIR` | `./data/agent-chat` (`/data/agent-chat` in Docker) | Local chat history, settings and empty working directory | In token mode, `/api` and `/mcp` require `Authorization: Bearer `; `/health` stays public. There is no user system — this is a local/self-hosted tool. > [!WARNING] > The defaults (`AUTH_MODE=none`, `HOST=0.0.0.0`) are meant for `localhost`. As shipped, > anyone who can reach the port can read **and modify** every workspace. Before exposing > StructSmith beyond your own machine, set `AUTH_MODE=token`, put TLS in front of it, and > publish the container port as `127.0.0.1:8090:8080`. See [SECURITY.md](SECURITY.md). --- ## Project layout ``` apps/ web/ React + Vite UI (React Flow, TanStack Router/Query, Zustand, shadcn-style UI) server/ Bun + Express (REST, MCP transport, SSE, static UI) docs/ Client setup and operating guides packages/ contracts/ Zod schemas and DTOs shared by frontend, backend and MCP domain/ Pure domain: rules, validation, operations, services (no Express, no SQLite) database/ Drizzle + bun:sqlite repositories implementing the domain ports mcp/ MCP server: tools, resources, prompts, Streamable HTTP handler migrations/ Plain SQL, applied in order at startup tests/ bun:test suite over the real domain and a real (in-memory) database ``` REST and MCP call the same services; only the `database` package knows SQL: ``` REST ──┐ ├──► Application / Domain services ──► Repositories ──► SQLite MCP ───┘ ``` ## Scripts ```bash bun run dev # API + UI, one command bun run build # production build of the UI bun run start # run the server with the built UI bun run test # domain + persistence tests bun run typecheck # strict TypeScript across the whole monorepo bun run db:generate # generate a migration with drizzle-kit bun run db:migrate # apply migrations bun run db:seed # (re)create the example workspace bun run mcp:stdio # MCP over stdio ``` ## Roadmap Systems and containers now offer **Open details** on their cards, in the inspector, and in the context menu. Double-click is a shortcut. A system opens its scoped container view; a container opens its scoped component view. Multiple matching views show a chooser. When none exists, an explicit creation dialog can seed a view with one level of children and their connected context, or start empty. **Back** and the navigation trail restore the previous zoom, pan, and selection within the current workspace session. Browsing existing views does not change the model. Ctrl/Cmd-click remains multi-selection. Use **View settings** beside the current view's name to edit its name in the inspector. Enter or leaving the field saves; Escape cancels. Renaming preserves the view's ID, links, scope, and layout, and supports undo/redo. The next improvements focus on dependency exploration, inline expansion, and visual overlays. Later milestones add message flows, offline presentations, visual change review, and saved proposals. See [the product roadmap](ROADMAP.md) for priorities, scope, completion criteria, and follow-up work including documentation, interoperability, and collaboration. ## Contributing Pull requests are welcome. [CONTRIBUTING.md](CONTRIBUTING.md) covers the local setup, the four commands CI runs, and the five architecture invariants a change has to respect. Participation is governed by the [Code of Conduct](CODE_OF_CONDUCT.md), and security issues go through [SECURITY.md](SECURITY.md) rather than public issues. ## License MIT — see [LICENSE](LICENSE). React Flow is used strictly within its open-source scope; the attribution is intentionally left in place.