Blurt skill
A skill and MCP for voiced screen recordings: the agent turns a spoken bug walkthrough into tasks with screenshots and code
- 283 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The app needs the microphone, screen recording, accessibility and input monitoring, and before dictation it reads about 500 characters around the cursor
- The transcript and screen frames reach the agent's model provider, and they can contain passwords, customer data and other people's chats
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/blurtDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The app needs the microphone, screen recording, accessibility and input monitoring, and before dictation it reads about 500 characters around the cursor
- The transcript and screen frames reach the agent's model provider, and they can contain passwords, customer data and other people's chats
This is third-party code. Review the repository files before installing.
What it does
The `blurt` skill collects feedback out loud. You record your screen and say what you see: demo the product, hunt bugs, discuss an idea. The agent transcribes speech locally, picks frames and turns the recording into a list. It holds bugs and polish issues with reproduction steps, ideas with context, notes and tasks. Each item gets a screenshot, a mark on it and the code that is probably involved. An example from the README: a five-minute walkthrough produced 52 items. Items are reviewed on a local page, then exported to Markdown or CSV, packed into a zip for a teammate or sent to any tracker and chat through the agent's own tools. By design Blurt has no tracker integrations. The repository installs the skill, an MCP server and a plugin for Claude Code. The MCP server provides the tools `search_items`, `get_item`, `get_transcript`, `update_item`, `mark_fixed` and `bundle_items`, so the agent reports work status back into the list itself. The skill comes with the Blurt macOS menu-bar app. The `Fn` key starts dictation in any app, `⌥A` snaps an area with a note, `⌥C` copies text from the screen and `⌥R` records the screen. Speech recognition runs on your computer.
Who it is for. Developers, testers and product managers who find it easier to talk through remarks about a product than to type each task into a tracker.
Good fit when
- Before a release you want to walk through the app and complain about every rough edge without stopping to file tasks
- A designer or client sends a video with comments and you need separate tasks with screenshots from it
- A terminal shows an error that is easier to show with a marked snap and a voice description than in text
- Text on the screen cannot be selected, for example in a screenshot, a video or a terminal, and it needs copying
- A list of remarks must be packed into one archive and handed to a teammate without access to the code
Not a fit when
- You need recording on Windows or Linux: the app is macOS-only for now, and elsewhere only the skill's scripts work, with Linux on X11 only
- The screen shows customer data or passwords, and recordings and snaps may reach your agent's model provider
- A single line in a tracker is enough: a screen recording is overkill for such a task
Example request
start blurt. I will record my screen and talk through what is wrong on the checkout page, then turn the recording into tasks and open the review pageLimitations
- The app needs macOS 13 or newer. Dictation and the recommended speech models need Apple silicon.
- Permissions for the microphone, accessibility, input monitoring and screen recording are required. On first run the skill builds the app, which needs the Xcode Command Line Tools and `uv`.
- Local models weigh from 228 MB to 3.4 GB and download on a button press. Cloud keys for Groq, OpenAI and DashScope are added if you want them.
- The transcript and a few screenshots of the recording go to your agent's provider, like any data you give it.
How to disable. Delete the Blurt app from Applications and the `~/Blurt` directory with recordings. Disable the plugin through `/plugin` and remove the skill by deleting the `blurt` folder from the skills directory. In projects delete the `.blurt/` folder as well.
FAQ
How do I install Blurt in Claude Code?
Run `/plugin marketplace add AGIHunt/blurt`, then `/plugin install blurt@blurt` and tell the agent "start blurt". For other agents `npx skills add AGIHunt/blurt -g` works.
How do I connect Blurt to Codex, Cursor and other agents?
In the Blurt app open Settings, the Agents section, and press Connect: it adds the skill and MCP server to the agents it finds after showing the changes. The list includes Claude Code, Codex, Gemini CLI, Kimi Code, OpenCode, Cursor and others.
Does Blurt work without the app?
The skill works without it: `record.py` records screen and sound, and transcription runs locally. macOS needs the Xcode Command Line Tools, and on Windows you install `uv` and FFmpeg.
How much does processing a recording cost?
The app itself is free, with no word limit and no account. Turning a recording into items spends your agent's tokens: in the README examples recordings of 7 to 20 minutes cost about 2 to 2.6 dollars at API prices.
Where does my data go?
Speech is recognized on the computer and the video is not sent to a model. The agent reads the transcript and looks at a few frames, so text and screenshots reach your agent's provider.
Discussion
No questions or reviews yet.
Related
SkillsA skill and MCP server let an agent search and analyze meeting, lecture and stream transcripts kept by the DeLive app
A macOS app that remembers your screen and opens the history to agents over MCP: search, days, work threads, screenshots.
A skill that gives an agent a pointer on your Mac screen, using arrows, boxes and labels to show the button you need.
About a hundred skills for Claude Code: meetings, research, publishing, TDD, Telegram, email and personal analytics
Local screen and audio recording with natural language search, feeding your activity history to an agent via MCP
MeetCap records Mac meetings and transcribes them locally; its skill lets Claude Code and Codex record and search the archive
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Blurt (this card) | no | 283 | ||
| VoiceMode | yes | not required | 1,439 | an MCP server for voice conversations with an agent; Blurt turns a screen recording with narration into tasks and types dictation into any app |
| Alice Voice for Hermes | no | 11 | voice input through Yandex Alice into a personal Telegram dialog for a single operator; Blurt runs on a Mac with local recognition and screen recording |
README in short
The page tells the idea of a keyboard for the agent era, describes four modes and app installation, then shows how the app, the skill and the MCP server fit together. Next come a comparison with other dictation apps, a privacy section and answers to common questions.
README badge
Original SKILL.md
---
name: blurt
description: >-
Screen + voice → structured work. The user records their screen while talking (demoing their product, browsing,
exploring an idea); turn the recording into items — bug/polish issues with repro steps, frames and suspected code,
ideas with the inspiration behind them, notes, tasks — let them review on a local page, then export (Markdown/CSV,
a hand-off bundle (.zip), or any tracker or chat through the tools you already have) or start fixing. Use when the user
wants to record feedback / QA / ideas by talking while using their screen, hands over such a video, wants to
process recordings from the Blurt app, or wants their Blurt items packed up and sent somewhere.
Triggers: "blurt", "/blurt", "start blurt", "开始口喷", "口喷鸡", "开始录", "录屏提 bug", "录一下我的想法", "record feedback",
"turn this recording into issues", "处理我录的视频", "process my recordings", "把这些问题打包发给…",
"bundle these items for…", "set up Blurt's models", "帮我设置口喷鸡的模型", "配置听写模型 / API Key".
---
# blurt — show it, say it, get structured work back
People think best out loud with their eyes on the screen. They jump between topics, correct themselves, point at
things. Your job: capture that with no friction, then do the tedious part (splitting, writing up, picking and
marking screenshots, finding the code, filing) with care. Optimise for the user's flow: never interrupt a
recording, ask few questions, make review fast.
Scripts live in `scripts/` next to this file; run them with `uv run <skill-dir>/scripts/<name>.py ...` (each
declares its deps; `uv` installs them on first use). Every script has `--help`. Always reply in the user's
language; write items in the language the user spoke.
## 0. Setup (first run, or when something fails)
- `doctor.py` → OS, RAM/GPU, ffmpeg, configured ASR, recommendations, `todo` list.
- Missing ffmpeg/uv: offer the install command; run it only after the user agrees.
- ASR: if none configured, pick from `asr_recommendations` + the language the user speaks, tell the user in one
line what and why (size, local vs cloud, cost), run its `setup`. Details: `reference/asr.md`. Cloud keys are set
by the user in their env — never ask them to paste keys in chat.
- `record.py build` — macOS: compiles the native recorder (Xcode Command Line Tools; else a Tk fallback) and
installs the **Blurt** menu-bar app into ~/Applications. Windows: the Start-menu shortcut appears on first
recording. Mention it in one line: "Blurt is also in your Applications — ⌥R records from anywhere".
- `record.py test` → look at the returned `frame` yourself. Wallpaper-only/black or `ok: false` ⇒ Screen Recording
permission missing; `mic_silent: true` ⇒ Microphone permission / wrong mic. On macOS the permission belongs to the
app hosting you (Terminal, iTerm, VS Code, Claude, Codex…) and that app must be restarted after granting.
macOS 15+ may show an "allow … to bypass the window picker" prompt — the user clicks Allow themselves.
- Optional project config `.blurt/config.json`: export target, owners/module map, column mapping, custom lenses in
`.blurt/lenses/`. `.blurt/` is never committed: items.py / projects.py add it to `.git/info/exclude` (not
`.gitignore`) the first time they write there.
## 1. Record (or pick up a recording)
**Just "/blurt" or "blurt", nothing else:** don't start a recording. Say where things stand in one or two lines —
the current project (`projects.py current`), new items (`items.py list --state new`), recordings not organized yet
(`record.py inbox`) — and offer three choices: record now, organize the new recordings, work on the items (fix or
export). Start a recording right away only when the user asked for one ("开始录", "录屏提 bug", "record feedback",
"start blurt").