PinchTab
@pinchtab
About PinchTab
High-performance browser automation bridge and multi-instance orchestrator with advanced stealth injection and real-time dashboard.
Config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"pinchtab": {
"command": "docker",
"args": [
"run",
"-d",
"\\"
]
}
}
}Tools
No tools detected
We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.
Overview
What is PinchTab?
PinchTab is a standalone HTTP server that gives AI agents direct control over Chrome. It is a 12MB Go binary with an SMCP (Model Context Protocol) plugin, providing token-efficient browser automation via HTTP API or CLI.
How to use PinchTab?
Install via curl, npm, or Docker, then run pinchtab to start the server on http://localhost:9867. Control the browser using CLI commands (e.g., pinchtab nav, pinchtab click) or the HTTP API directly. For MCP integration, configure the SMCP plugin (see plugins/README.md for environment variables and paths).
Key features of PinchTab
- Token-efficient text extraction (~800 tokens per page)
- Headless or headed browser modes
- Multi-instance with isolated profiles
- Self-contained 12MB binary, no external dependencies
- Accessibility-first with stable element references
- ARM64-optimized including Raspberry Pi support
Use cases of PinchTab
- AI agent automation: navigate, click, fill forms, and press keys
- Token-efficient data extraction from web pages
- Multi-instance workflows with parallel isolated profiles
- Persistent browser sessions with profile reuse across restarts
FAQ from PinchTab
What is the difference between PinchTab server and bridge?
The server is the main control-plane process that manages profiles, instances, and routing. The bridge is a thin per-instance runtime used internally for managed child instances; most users only need the server.
How does PinchTab achieve token efficiency?
PinchTab extracts page text at ~800 tokens per page instead of using screenshots, which can be 5–13x cheaper in token usage.
What security measures does PinchTab include?
PinchTab defaults to local-first with server.bind = 127.0.0.1. It uses an API token for access control and IDPI (Input Domain Protection) with a local-only website allowlist as a defense layer against prompt injection.
Can PinchTab attach an externally managed Chrome instance?
Yes, the "attach" feature allows registering an externally managed Chrome process, but it is disabled by default for security.
Is PinchTab compatible with MCP?
Yes. The repository includes an SMCP plugin that exposes 15 tools (e.g., pinchtab__navigate, pinchtab__snapshot) for use with the Model Context Protocol.
Frequently asked questions
What is the difference between PinchTab server and bridge?
The server is the main control-plane process that manages profiles, instances, and routing. The bridge is a thin per-instance runtime used internally for managed child instances; most users only need the server.
How does PinchTab achieve token efficiency?
PinchTab extracts page text at ~800 tokens per page instead of using screenshots, which can be 5–13x cheaper in token usage.
What security measures does PinchTab include?
PinchTab defaults to local-first with `server.bind = 127.0.0.1`. It uses an API token for access control and IDPI (Input Domain Protection) with a local-only website allowlist as a defense layer against prompt injection.
Can PinchTab attach an externally managed Chrome instance?
Yes, the "attach" feature allows registering an externally managed Chrome process, but it is disabled by default for security.
Is PinchTab compatible with MCP?
Yes. The repository includes an SMCP plugin that exposes 15 tools (e.g., `pinchtab__navigate`, `pinchtab__snapshot`) for use with the Model Context Protocol.
Basic information
More Productivity MCP servers

YouTube Transcript AI
youtube-transcript.aiFetch the full transcript of any YouTube video as clean text — no API key, no signup, no quota wall. Pass a YouTube URL (watch, youtu.be, shorts, embed) or a bare video ID and get the transcript back with title and metad

AdminLanding
Julien MauriceAdminLanding — French rental, frontalier & home-employment compliance tools (FR/EN). Sourced, dated answers from deterministic rule engines and official datasets — no AI, no account needed for the 10 free tools: zone ten

My AskAI
My AskAIUse My AskAI to find product, setup, integration, API, feature, pricing, and troubleshooting information. Search the public knowledge base, read complete documentation and OpenAPI files, and estimate pricing from monthly

LinkedIn Ghostwriter
ContentInWrite LinkedIn posts in your own voice from Claude, ChatGPT, or any MCP client. ContentIn trains a personal AI on your own LinkedIn posts, then generates ideas, drafts in your voice, schedules, publishes via LinkedIn's o

Prose Coach
prose.coachDeterministic writing filter. 43 AI patterns marked, each quoting the line that triggered it and attaching the fix. Connects as a remote MCP server at one URL. The free tier runs 3 drafts a day at 12,000 characters, no a
Comments