█

AI/TLDR

Whiteboard

An open-source canvas where coding agents draw diagrams of their changes, so people review the design before the diff

AI Code Review & SecurityOpen source
Latest
v0.1.2
Updated
24 Sep 2026
Language
TypeScript
License
MIT
Coverage
1 story

What's new

v0.1.224 Sep 2026

Public launch on Show HN, with the v0.1.x desktop releases for macOS and Fedora: an agent SDK for drawing diagrams on the canvas, a Rust semantic diff viewer and a Decision Log for auditing agent choices.

Latest news

Overview

Whiteboard is an open-source desktop app, built by the Y Combinator W26 startup /dev/fast, where humans and coding agents architect software together in a shared workspace. It does not replace your coding agent. It plugs into the ones you already use, such as Claude Code and Codex, and gives the agent an SDK to draw on an in-app canvas that explains its work: sequence diagrams, entity-relationship diagrams and other architecture-level views.

Every element on the canvas links back to the code behind it. The app vendors Code OSS as its editor, so navigation uses VS Code keybindings and LSP support. Next to the diagrams sits a semantic diff viewer written in Rust: it compares code by its syntax tree, can show long functions as pseudocode, always shows function signatures, and folds tests and documentation by default so the reviewer starts with the parts that matter.

A Decision Log lets agents query and link their own traces, so a reviewer can see the requirements, the implementation choices and the decisions the agent made on its own, and check whether they make sense. Whiteboard runs locally against your repositories on macOS and Fedora Linux and is MIT-licensed. It sends anonymous telemetry that excludes code, diffs, prompts and model output, and it can be switched off. The README lists current limits: no file editing inside the app, one repository per review, and shared reviews that do not update across machines.

What it does

  • Agent SDK that lets Claude Code, Codex and other agents draw on an in-app canvas to describe their work
  • Sequence diagrams and entity-relationship diagrams that link each element to the underlying code
  • Code OSS editor with VS Code keybindings and LSP navigation
  • AST-aware semantic diff viewer in Rust that can show long functions as pseudocode and folds tests and docs by default
  • Decision Log for auditing the requirements, implementation choices and autonomous decisions in an agent's trajectory
  • Runs locally against your repositories; telemetry excludes code, diffs, prompts and model output and can be disabled

Getting started

Whiteboard is a desktop app for macOS and Fedora Linux. Install it from the project's download links, then ask the coding agent you already use to write a Whiteboard review of its change.

Videowhiteboard demo (yc w26) - ide for thoughtful softwaredevfast (yc w26) ↗

Download the app

The README links a macOS installer at install.dev.fast and a Fedora build at install.dev.fast/linux.

texttext
https://install.dev.fast
https://install.dev.fast/linux

Open a repository

Whiteboard runs locally against your code repositories. Each review covers one repository at a time.

Have your agent draw a review

From your terminal agent (for example Claude Code or Codex), ask it to write a Whiteboard review of its change. The agent uses Whiteboard's SDK to draw diagrams on the canvas, which you then read alongside the semantic diff and the Decision Log.

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

When to use it

  • Reach for it when an agent's change is too large to read line by line and you want to check the architecture first
  • Reach for it when reviewing spec-level or data-model changes, where a sequence or ER diagram says more than a diff
  • Reach for it when you need to audit which decisions an agent made on its own during a long run
  • Reach for it when you want a diff that hides noise like tests and docs and shows long functions as pseudocode

How Whiteboard compares

Whiteboard alongside other open-source ai code review & security tools AI/TLDR tracks, ranked by GitHub stars.

ToolStarsWhat it does
Open Code Review★ 41kAlibaba's code-review CLI that pins file selection, bundling and rule matching in deterministic pipelines and leaves only judgement to an LLM agent, producing line-level comments.
Bend★ 22.7kCompiled language whose type checker doubles as a proof checker, so an edit only builds if it can be proved to keep the rules declared in a LAWS.bend file.
Codex Security★ 10.8kOpenAI's CLI and TypeScript SDK that scans a repository for security vulnerabilities, deduplicates findings by embedding similarity and re-grades their severity against your own policy.
CodeRabbit—AI pull request reviewer that analyzes diffs with full-repo context to flag bugs, security issues, and quality problems, posting inline comments and one-click fixes.
Greptile—AI code review tool that indexes your whole codebase into a graph so a swarm of agents can catch multi-file logic bugs and security risks in every PR.
Qodo—Agentic code-quality platform whose Qodo Merge reviews PRs with context-aware suggestions, test generation, and team-standard enforcement across Git hosts and IDEs.
Graphite—Stacked-PR developer workflow platform whose Diamond AI reviewer gives high-signal pre-merge feedback and suggested fixes on every pull request.
Whiteboard—An open-source canvas where coding agents draw diagrams of their changes, so people review the design before the diff