Whiteboard and Radix Debut as Interfaces for Agentic Software Development

Whiteboard is being built by Sid, Alex, Ketan, and Milan, four college friends who say they left tech-lead jobs shortly before agentic coding became standard. The app plugs into tools such as Claude Code and Codex and gives an agent an SDK to draw on an in-app canvas to describe its work, according to the team’s Hacker News post. The project began as an MVP based on HTML artifacts, but the team reworked it after running into limitations.

Built on CodeOSS, Whiteboard lets users click visualizations—sequence diagrams, entity relationship diagrams, or quotes from an agent’s trace—and jump to the underlying code, the team said. It provides VSCode keybindings and LSP support. The team wrote a semantic, AST-aware diff viewer in Rust that shows only relevant code changes; by default, large added functions are summarized as pseudocode, while unit tests and large documentation changes are collapsed or hidden. A WASM-based plugin system allows customization. Whiteboard also includes a Decision Log, with tools for agents to query and link their own traces so users can see how requirements were implemented and what decisions the agent made autonomously.

Salesforce and Modal employees use Whiteboard as a review tool for architecture or spec-level changes, according to the post. The team cited two main uses: reviewing a coding agent’s own work, often by having the agent create a prototype and a corresponding Whiteboard session for design iteration, and reviewing other people’s changes, especially when composed with tools like Greptile. In that workflow, an automated reviewer can handle small changes and escalate changes requiring human judgment to a Whiteboard session.

Whiteboard is released under the MIT license for macOS and Linux. The team said it eventually plans to charge companies for a hosted web version that manages whiteboard session creation alongside trajectory storage and multiplayer reviews, while keeping everything self-hostable.

Radix, announced in a separate Hacker News Show HN post by Jordan, is a UI tool for programming agents. A user prompts an agent to generate a workspace for a task and receives an interactive widget that persists locally on disk. Jordan said he built Radix because he often needed tools while writing code—for testing small experiments, trying UI ideas, or visualizing results—and found existing options, including Claude artifacts, limited.

He wanted a system where users could leave comments or adapt an artifact without giving vague descriptions back to an agent, and ideally move away from a chat window as the main interface, according to the post. Radix is currently free and in beta, with a pro version planned. Jordan said there is no telemetry or data capture, messages run entirely through the user’s own agent, and a key is required only to gauge usage. Everything is stored locally, and the workspaces are React apps that can be edited manually.

Both projects address a shift in software development toward agentic coding, where AI agents generate code and other artifacts at increasing speed. Whiteboard’s team described the problem of cognitive debt when more pull requests are merged without developers understanding them, making it difficult to contribute to a system. Whiteboard’s answer is a common visual workspace tied directly to code and agent traces; Radix’s answer is a persistent, local widget layer that gives agent-generated tools a more interactive form than a chat response.