molerat

An accessibility & agentic layer for Rust UI applications, built on AccessKit 🐀

Quickstart

cargo add molerat

Add molerat to your app and publish a semantic view alongside your existing UI:

use molerat::Molerat;

let mut molerat = Molerat::new("Notes");
molerat.publish(|ui| {
    ui.text("status", "All changes saved");
});

See the available semantic UI components in the API documentation.

If you have actions in your app, define an enum and return them from your UI handlers:

#[derive(Debug)]
enum Message {
    Save,
}

let mut molerat = Molerat::<Message>::new("Notes");
molerat.publish(|ui| {
    ui.button("save", "Save").on_click(|| Message::Save);
});

See the available semantic actions in the API documentation.

To handle the incoming actions:

for message in molerat.messages() {
    match message {
        Message::Save => { /* update app state */ }
    }
}

Examples

• ratatui-recipe-atspi: An app that speaks Ratatouille recipe.
• ratatui-dashboard-atspi: An app that speaks the system stats.
• ratatui-recipe-mcp: Task queue controlled by agents.
• ratatui-sqlite-mcp: SQLite workbench controlled by agents.
• ratatui-paint-mcp: Drawing canvas controlled by agents.

Run one with:

cargo run --example <example-name> --features <mcp|atspi>

# For example:
cargo run --example ratatui-recipe-atspi --features atspi

# The MCP examples require the stdio bridge (molerat-mcp) to be installed:
cargo install molerat --features mcp --bin molerat-mcp
cargo run --example ratatui-recipe-mcp --features mcp

The examples above are built with Ratatui, but molerat is not designed for any specific UI framework.

[!TIP]
If you want to integrate molerat with Ratatui, see the migration guide or simply give that document to your agent 🧀

Showcase

Here is knife (a reverse-engineering TUI) solving a crackme challenge:

https://github.com/user-attachments/assets/586710de-214c-468c-91b7-0126d7a25592

Here is BugStalker (a Rust debugger) debugging an application:

https://github.com/user-attachments/assets/ff256f9a-4f7b-4374-a7d3-5d91e2ebec55

Integrations

Screen readers

You can use the AT-SPI to expose your UI/actions to screen readers on Linux!

1. Enable atspi feature:

cargo add molerat --features atspi

1. And then adjust your constructor to use AT-SPI:

let mut molerat = Molerat::<Message>::new_atspi("Notes");

Nothing else is needed!

Agents

You can use the Model Context Protocol (MCP) to expose your UI/actions to agents!

1. Install the molerat-mcp stdio bridge and register it with your MCP client:

cargo install molerat --features mcp --bin molerat-mcp

1. Register the installed binary with your MCP client. e.g. Codex:

codex mcp add molerat -- "$(command -v molerat-mcp)"

1. Enable mcp feature:

cargo add molerat --features mcp

1. And then you can change your constructor to use the MCP bridge:

let mut molerat = Molerat::<Message>::new_mcp("Notes")?;

That's it! Your app is now agentic and can be controlled by your favorite agent.

License

Licensed under either of Apache License Version 2.0 or The MIT License at your option.

🦀 ノ( º _ º ノ) - respect crables!

Copyright

Copyright © 2026, Orhun Parmaksız