v1.0.0 — Now Available

Record it.
Agents see it.

The free, open-source screen recording CLI for AI agents. Claude Code, Codex, Cursor, Z Code — they all see what you see. Zero cost. 100% local.

Get Started View on GitHub
claude mcp add recit -- npx -y @zum-x/recit
$0
Total Cost
0
API Keys
3
Platforms
5
MCP Tools
100%
Local & Private
Features

Everything agents need to see your screen

Video recording, screenshots, frame extraction, event logging — all through a simple CLI that any AI agent can call.

Screen Recording

Record full screen, specific windows, or regions. ffmpeg-powered, cross-platform. Default 15fps, configurable up to 60fps.

Free forever

Instant Screenshots

Capture the current screen in a single command. Outputs PNG with metadata. JSON mode for structured agent consumption.

Free forever

MCP Server

5 tools exposed via Model Context Protocol. Works with Claude Code, Cursor, Codex, Windsurf, Cline — any MCP client.

stdio — no server

Frame Extraction

Extract key frames from recordings using scene detection. Token-aware output for vision model budgets.

Free forever

Event Logging

Log mouse clicks, keystrokes, window switches synced to video clock. JSONL format. Perfect for agent replay.

Free forever

Before/After Diff

Visual comparison output for agent reasoning. Essential for UI testing loops — "what changed after my code edit?"

Free forever
How It Works

One command. Agent gets eyes.

Your AI agent spawns recit as a local process. No cloud, no hosting, no server. Communication over stdin/stdout. That's it.

Claude Code
Codex
Cursor
Z Code
Windsurf
Cline
# 1. Install (one-time) claude mcp add recit -- npx -y @zum-x/recit # 2. Agent captures screenshot Agent calls: recit_snap --json Returns: { "path": "/tmp/recit_snap_...", "size": 2048, "format": "png" } # 3. Record a full session Agent calls: recit_start --output ./demo.mp4 Agent does work... Agent calls: recit_stop rec_123 Returns: { "path": "./demo.mp4", "duration": 12.5, "size": 307200 } # 4. Extract frames Agent calls: recit_frames ./demo.mp4 --max 10 Returns: { "frames": [...], "count": 8 }
AI agent recording screen
Agent Vision

Give every AI agent a screen recorder

Your agents can now watch what you watch. Record a bug, capture a workflow, document a process — the agent sees it and acts on it. No manual screenshots needed.

Screen recording for AI agents
Zero Cost Architecture

ffmpeg + MCP + stdio = $0 forever

No cloud APIs. No paid services. No hosting. ffmpeg handles recording. MCP handles agent integration. stdio means it runs on the user's own machine.

Get Started Free
Terminal CLI interface
CLI-First

Pipe it. Script it. Agent it.

recit is a Unix citizen. JSON output for agents. Human-readable stderr for you. Works in Docker, CI/CD, SSH sessions, and your local terminal.

Quick Start

Up and running in 60 seconds

Three steps. Zero configuration. No accounts.

Install

One command adds recit to your agent's toolkit. No download, no setup, no API keys.

Record

Tell your agent to record. It calls recit via MCP. ffmpeg captures your screen locally.

Analyze

The agent reads the video, extracts frames, understands what happened, and takes action.

Works With

Every major AI agent, out of the box

Because recit uses the standard Model Context Protocol, it works with every agent that supports MCP. No custom integrations needed.

Agent Setup Status
Claude Code claude mcp add recit -- npx -y @zum-x/recit Native
OpenAI Codex Shell tool → npx -y @zum-x/recit Works
Cursor .cursor/mcp.json config Native
Z Code MCP config or shell Works
Windsurf Cascade MCP config Native
Any MCP Client Standard MCP protocol Universal
Install

One line. Done.

No npm install needed. npx downloads and runs recit on demand. You just need Node.js 18+ and ffmpeg (both free).

For Claude Code
claude mcp add recit -- npx -y @zum-x/recit
For Cursor / VS Code
{ "mcpServers": { "recit": { "command": "npx", "args": ["-y", "recit"] } } }
Prerequisites
# Both free, both common node --version # v18.0.0+ ffmpeg -version # any version # Install ffmpeg if missing: brew install ffmpeg # macOS sudo apt install ffmpeg # Linux winget install ffmpeg # Windows
Open Source

Free forever. Apache 2.0 licensed. No catch.

Built with ffmpeg + TypeScript + MCP. Published to npm. Source on GitHub. Your data never leaves your machine.

Star on GitHub Install Now