From idea to release, one agent works with the whole team.

Des2Dev is a desktop app for building a product end to end with an AI agent. The brief, design, spec, code, testcases and release notes live in one repo, linked to each other and in step with Figma.

Start free

Free. Uses the agent plan you already have.

acme-appVersion 12 · approvedPush

Add a Sign out button to Settings, under the profile card.

Readdesign/figma/screens/settings.ts
Editeddesign/figma/screens/settings.ts
des2dev check12 screens, no problems

Sign out is in Settings, under the profile card, in Light and Dark. The check passes.

✓ Done in 41s

Ask Claude to change a screenOpus

PreviewNotes

Sign in

Dashboard

Settings

1

Five tools, five versions of the truth.

The PM writes in one place, the designer draws in another, developers code in a third and QA tests in a fourth. Every handoff pulls the product a little further apart.

A tool for every role

The brief in Docs, the design in Figma, tasks in Jira, testcases in a spreadsheet. None of them knows about the others.

Documents go stale

The code changes and nobody updates the spec or Figma. By test time, nobody is sure which version is right.

Handoffs drop details

PM to designer, designer to developer, developer to QA. Each handoff is another chance to get it wrong.

AI work gets lost

The agent codes in a chat. Its reasons, its decisions and what it made scroll away with the conversation.

With Des2Dev, one flow from start to finish.

You decide, the agent writes. Every step leaves documents in the repo for the next step to build on.

Requirement
R2Sign out
As a signed in user, I want to sign out from Settings, so nobody else uses my account on this device.
- R2.1WHEN the user taps Sign out THEN the session ends
- R2.2WHEN the session ends THEN the app opens Sign in
01 · Start
Shape the product
You
Describe the idea: the problem, who it is for, how you will measure success.
The agent
Writes the product brief, requirements with acceptance criteria, the user flow and the sitemap.
You get
Product briefRequirementsUser flowSitemap
02 · Middle
Design and build
You
Review each screen and comment on the canvas or in Figma.
The agent
Draws screens from the design system, writes the tech spec, API and data model, splits the work into tasks and codes them.
You get
ScreensDesign systemTech specAPITasks
03 · End
Test and release
You
Run the tests, read the results, push the new version.
The agent
Writes testcases from the acceptance criteria, checks accessibility, files bugs, drafts release notes and docs.
You get
TestcasesBugsAccessibilityRelease notesDocs

ID	Case	Covers	Priority
TC-1	Sign out ends the session	R2.1	P0 · e2e
TC-2	Sign out offline	R2.1	P1 · manual
TC-3	Lands on Sign in	R2.2	P1 · e2e

At the end, the whole product is in one repo.

Code, design, spec, tests and their history sit side by side. Each one knows where it came from.

One source of truth

Every document is a file in the repo, reviewed in pull requests like code.

See what is missing

Coverage points out criteria without tests, screens without a requirement and strings without a translation.

Figma matches the code

Every push redraws the screens and the design system in Figma. Designers keep working there.

History for everything

Every document has its own versions. Compare two, restore an old one.

Every change goes all the way round.

When a requirement changes, the spec, screens, testcases and tasks change with it, in the same pass.

A requirement changes

Edit a requirement and the agent updates the screens, testcases and tasks tied to it.

The designer comments

Notes from Figma or the canvas land in the chat. The agent fixes, the designer approves.

A test fails

A failed testcase becomes a bug on its screen. The agent fixes it and reruns that case.

Before and after Des2Dev.

The same team and the same product, built a different way.

	The old way	With Des2Dev
Where documents live	Docs, Figma, Jira, spreadsheets, each apart	One repo, all linked
When a requirement changes	Edited by hand in each place, often missed	The agent updates everything tied to it
Design and code	Drift apart within a few sprints	Figma redrawn from code on every push
Testcases	Written late, incomplete, unlinked	Written from acceptance criteria, with coverage
Handoffs	Meetings, screenshots, messages	The whole team opens the same repo
What the AI makes	Stays in a chat and gets lost	Saved as files in the repo, with versions

Chat on one side, the project on the other.

The desktop app runs Claude Code. What the agent writes opens next to the chat.

Chat with Claude Code

Your own Claude Code works from the chat. Reopen a chat to pick up where it stopped.

Document list

Brief, screens, diagrams, spec, tests, tasks and the rest, each with a count.

Screen canvas

Every screen in the design, Light and Dark, redrawn as files change.

Ask from a document

Click a requirement, task or screen to put a ready request in the chat.

Push with a preview

Claude cannot push. You read the check and the diff from the last version, then push.

You approve risky steps

Shell commands and web fetches wait for Allow, Always allow or Deny.

Use the agent you already have.

Des2Dev runs no model of its own. Your agent does the work, on your plan.

Desktop app

Chat with Claude Code inside Des2Dev.

1Claude Code installed and signed in
2Model usage goes on your Claude plan or API
3Fable, Opus, Sonnet or Haiku, and the effort, per chat

Your agent over MCP

Every des2dev command as a tool.

1des2dev mcp install <agent>
2Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Kiro, OpenCode
3Checks, pushes, reads notes, builds the approved version

Terminal

The same commands, typed by hand or run in CI.

1des2dev init in the repo
2des2dev push, pull, feedback, approved
3No agent needed

Questions

Open a repo. Tell it your first idea.

Sign in, run des2dev init in your repo, and tell your agent what you want to build.