Show HN: Whiteboard (YC W26) – An open-source IDE for thoughtful software design
Whiteboard is an open-source (MIT) desktop IDE that gives coding agents like Claude Code and Codex an SDK to draw diagrams, sequence charts, and traces on a shared canvas linked directly to underlying code. It vendors Code OSS for VSCode-grade navigation/LSP support and adds a custom Rust-built AST-aware semantic diff viewer (with a WASM plugin system) that summarizes large generated functions as pseudocode and collapses noise like tests/docs, aiming to make reviewing autonomous agent decisions and code changes tractable.
What it claims
Whiteboard is a desktop review app (not really a full IDE for editing — see Limits) that gives your existing coding agent an SDK to draw sequence diagrams, ER diagrams, and trace links on a canvas, with clickable jumps from any visualization back to the underlying code. The summary covers the architecture; the page adds the operational envelope, the recommended models, and several hard limitations the summary omits.
How to use it
Distribution is narrow: downloads are offered for macOS and Fedora only — no general Windows or Linux build is listed. The workflow:
- Download and open the app.
- Connect Claude Code, Codex, or another agent from the welcome screen.
- Ask the agent to review your current branch against up-to-date main and open the result.
The authors say it works best with GPT-6 Sol and Claude Opus 5.5 for the intelligence/cost/speed tradeoff. It runs against your local checkouts. Feedback loop: highlight something you dislike, paste it to your agent, and it redraws.
Limits
These are stated on the page and materially affect whether this fits your workflow:
- You cannot edit files in Whiteboard. It is a review/design surface, not an editor — despite the "IDE" framing in the title.
- Multi-repo reviews aren't well supported — working across repos in a single review is not handled.
- Shared reviews don't sync. You can share a review between machines with the share button, but updates made after sharing don't propagate; you must re-share.
- Telemetry is anonymous and excludes your code, diffs, prompts, and model output, and can be turned off.
On the Code OSS approach
The page explains why it vendors Code OSS rather than maintaining patches (as most forks do): agents struggle with patches, and ~45% of stock VS Code is Copilot code they don't need. They monitor upstream and merge security/feature patches as they arrive. The vendored fork retains Microsoft's MIT license; see apps/review-desktop/LICENSE and apps/review-desktop/UPSTREAM.
A hosted product for teams is planned; the project states everything will always remain self-hostable.
Organisations named here
- Y Combinator
- Microsoft
- Claude Code
- YC W26
Technologies named here
- GPT-6 Sol
- Whiteboard
- Claude Code
- Claude Opus 5.5
- Codex
- Code OSS
- WASM-based plugin system
- Rust
- VS Code
- WASM
Techniques named here
- Semantic, AST-aware diffing
- Agent trace linking
Related in coding-agents
