antithesishq/bombadil

What it does

Bombadil is a property-based testing tool for web and terminal user interfaces. Instead of replaying fixed scripts, it autonomously explores the UI and validates correctness properties, aiming to surface harder bugs earlier. It runs in a local developer environment, in CI, and inside Antithesis. The project is written in Rust and is distributed under the MIT license.

Quick start

Bombadil's README does not contain installation steps or a runnable example; it points readers to the manual. Start from the Bombadil Manual, follow the installation instructions for your platform, then study the examples to see how a specification is written. A specification defines action generators and properties; the default specification includes a set of action generators that are picked from randomly, and custom specs can define their own action generators and weights (maintainer reply).

Usage constraints

The tool needs a UI to drive, so the target web or terminal application must be reachable from the machine running the test. Local runs inside Antithesis behave differently from ordinary environments: reproduction of failures is described as shaky unless the app is deterministic or runs inside Antithesis (maintainer reply). There is no built-in rerun support yet. Contributions are coordinated through GitHub issues and Discord rather than drive-by pull requests, and LLM-assisted contributions must follow the project's Gen AI Policy.

HotRepo's take

A March 19, 2026 Hacker News thread about the project drew 257 points and 106 comments (thread), and on November 15, 2026 the Rust daily list recorded it at position 6 with 189 stars gained that day. The discussion is more useful than the star count: one commenter praised the single-executable delivery and the absence of a heavy dependency chain (comment), which matters for a testing tool that must slot into existing CI. Others asked how actions are chosen and whether shrinking and rerunning exist, and the maintainer confirmed random default action generators and that reruns are not implemented yet (reply). The concerns raised about property-based UI testing are about the layers between inputs and the app under test (comment), which is worth weighing when deciding how much of a failure to attribute to your own code.