Design, build, document and review in one place. Bring your AI agents into the process, keep every decision connected, and store it all in your own Git repository.
Creates an order and captures payment. Idempotent on Idempotency-Key.
Three surfaces, one contract
The canvas, VS Code and the MCP server all read and write the same files in .forward/, so nothing drifts between design, editor and agents.
Draw interfaces, model architecture, narrate flows, design data and track work in the browser, served locally from your repository.
Backend developers attach framework-agnostic API documentation to any design element without leaving the code.
49 tools with full parity, over MCP or straight from the shell, so any AI assistant or script can create and maintain every artifact.
The building blocks
Separate layers that each attach to the design or link to it by id, stored one file per record so they merge cleanly and diff readably in a pull request.
Frames, rectangles, ellipses, text, reusable components and instances, with absolute positioning, nesting and styling. Run it locally with forward serve or open it inside VS Code.
Attach framework-agnostic Markdown docs to any element, from VS Code, an AI agent or the canvas. Endpoints stay tied to the interface they power, with an append-only history of every edit.
Your architecture as a project-wide graph (frontends, backends, databases, caches, queues, services, DevOps and external dependencies) overlaid on the canvas.
Step-by-step walkthroughs rendered as numbered paths across the graph. Tie a step to an endpoint and render the flow as a Mermaid sequence diagram.
Entity-relationship diagrams with tables, fields and crow's-foot relationships, serialized to a Mermaid erDiagram.
Status, priority, type, assignees, dependencies, comments and an activity feed. Every task can link to a design element, a system flow or an API endpoint, so work always tracks against the thing it's about.
Documentation where it appears in the UI
Pick a design element, write Markdown next to your code, and attach it from the VS Code extension, or let an AI agent do it through the MCP server. Every API endpoint stays associated with the interface it powers.
Designs from frontend code
forward capture renders your running app and reads back what the browser actually painted: one frame per route, a box for everything that paints, text for every run of text. React dev builds also report components, so repeated ones become a component plus instances.
Architecture & narrated flows
Visualize frontend apps, backend services, databases, caches, queues and integrations as one project-wide graph. Then narrate a flow (registration, auth, checkout) as a numbered path across it, and render it as a Mermaid sequence diagram.
Data models
Design and maintain entity-relationship diagrams with tables, fields and crow's-foot relationships. They serialize to Mermaid erDiagram definitions, and each entity is its own JSON file, easy to review in a pull request and version with the schema it describes.
Never lose the thread
Every edit to an element's API docs is an append-only history entry with a reviewable, git-style diff, and the project timeline is built from your git log, so commit messages show up in it too.
Save named snapshots of the whole canvas (design and infrastructure) at milestones, each with a note and per-layer change counts. Restore any of them whenever you need to.
Pick a baseline, or a git ref with forward diff origin/main, and get the endpoints to integrate, modify and remove, plus a summary of everything else that changed.
AI-agent first
Claude Code, Codex, Cursor, Copilot or any MCP client gets the same tools as the canvas: documentation, architecture, data models, tasks and the design itself. Agents without MCP run the very same tools from the shell, and every write is stamped with your git name plus the agent that made it.
Git-native
There is no central server or database. Whoever can push to your repository can change the project, and changes reach the rest of the team the way code does: through commits, pull requests and merges. The contributor roster is read straight from git.
Why Forward
Modern teams keep designs, API docs, infrastructure, data models, tasks and AI workflows in disconnected tools, and as the project evolves they drift out of sync. Forward keeps every artifact connected to the interface it supports, in the repository it describes.
Design, documentation, architecture and tasks live in one .forward/ folder next to the code, so they stop drifting apart.
Documentation lives next to the element it describes and versions with it, change by change, in the same commits as the code.
Backend devs document straight from VS Code: no context switch, no second tool to keep open.
An MCP server and a CLI give any agent full parity to read elements and maintain every artifact.
Tasks link to design, flows and endpoints, so work always tracks against the thing it's about.
Git history, per-element doc history and checkpoints record every change, by whom and when, so decisions are never a mystery.
Open source
Forward is open source under the Apache 2.0 license. Your project is plain JSON in .forward/ in your own repository: there is nothing to host and nothing to lock you in.
Use it, fork it, change it. Every surface (canvas, CLI, MCP server, VS Code extension) lives in one repository.
One file per record in .forward/, validated against one schema. Merges stay clean and pull-request diffs stay readable.
No server to run and no database to back up. Git is the sync, the history and the access control.
Run this in the root of your repository. It creates .forward/, the agent configs and the GitHub workflow; commit them and your team, the canvas and your agents are in sync.
npx -y -p forward-mcp forward init