For developers

The dictation tool that respects your terminal

Most dictation tools treat every field the same and turn a shell command into a sentence. Mr.Speech knows what it is typing into, runs natively on Linux, and exposes your notes to the AI tools you already use.

  • Native Linux, Wayland and Hyprland
  • Leaves terminal syntax alone
  • Notes reachable from Cursor and Claude Code
  • Runs offline
For developers

How it fits the day

01

Linux is a first-class platform

Not a port and not an afterthought. Real Wayland, Hyprland and KDE support, with a `.deb` and a `.rpm`, alongside native macOS and Windows builds from the same codebase, written in Rust.

02

It knows a terminal from a document

Untouched syntax in a terminal, structured prose in an editor, a casual line with no trailing period in chat. Zero setup, because it detects the app you are dictating into.

03

Your notes are reachable from your tools

59 tools locally over the Model Context Protocol for Claude Desktop, Cursor and Claude Code, and 45 over a cloud connector for ChatGPT and Claude. They read and write, not just read.

04

Commit messages, comments and design docs

The parts of the job that are prose, dictated at speaking speed, straight into the editor. A knowledge graph and semantic search across everything you wrote down.

05

It works when the network does not

On-device speech, on-device semantic search, on-device meeting transcription. No account required for any of it.

For developers

Where it earns its place

Design

Think out loud, keep the result

Talk through an approach and end with a linked note rather than a lost train of thought.

Review

Long review comments

Say the reasoning once; it lands in the review box formatted for that box.

Agents

Your workspace as a data layer

Ask your assistant to pull yesterday's decisions into a board, and it can, because the connector writes.

Questions

Questions

Does it really work on Wayland?

Yes, including Hyprland and KDE, with the injection path chosen per environment rather than assuming X11.

Will it mangle code I dictate?

The app context detection is there precisely to stop that: in a terminal, syntax is left as spoken. There is also a literal cleanup style that does not restructure anything.

What is the MCP connector for?

It lets an external AI assistant read and write your notes, tasks and boards. So an assistant can create a note or move a card, instead of you copying text between two windows.

For developers

The dictation tool that respects your terminal

Most dictation tools treat every field the same and turn a shell command into a sentence. Mr.Speech knows what it is typing into, runs natively on Linux, and exposes your notes to the AI tools you already use.