How Pinchtab Works
Pinchtab operates as a self-contained 12MB Go binary that launches its own Chrome instance, either in a visible "headed" mode for debugging or, more commonly, an invisible "headless" mode for pure automation. It exposes a simple, plain HTTP API, allowing external AI agents to issue commands such as navigating to URLs, clicking buttons, typing text, or extracting page content. A core innovation is its "accessibility-first" approach, which generates a structured tree of the webpage. This tree provides stable references for elements (e.g., e0, e1), enabling agents to interact deterministically and reliably, without relying on error-prone visual analysis or XPath selectors. This method significantly reduces the number of tokens required to understand and interact with a page.
Why Use Pinchtab?
Unparalleled Flexibility and Agent Agnosticism: Pinchtab's HTTP API means it's not locked into any specific agent framework. Whether you're using OpenClaw, another custom AI agent, or even simple curl commands, Pinchtab seamlessly integrates, allowing developers to switch agents or script interactions in any language.
Cost-Effective Web Automation: By providing highly optimized content extraction (e.g., text endpoint for ~800 tokens per page) and interactive-only snapshots (~3,600 tokens), Pinchtab is 5 to 13 times cheaper than full snapshots or screenshots for read-heavy tasks. This token efficiency translates directly into lower operational costs for AI models.
Reliability and Stealth: Interactions are based on deterministic accessibility tree references, eliminating the guesswork associated with vision models. Furthermore, Pinchtab includes a robust "stealth mode" that bypasses common bot detection mechanisms by patching navigator.webdriver, spoofing user agents, and hiding automation flags.
Persistent Sessions and Zero Config: Pinchtab manages persistent browser profiles, allowing agents to log into sites once and maintain authenticated sessions across restarts. Its self-contained binary and zero-configuration setup make it incredibly easy to deploy and use, especially in Docker or CI environments.
Ideal User: Pinchtab is perfect for AI developers, researchers, and engineers who need a powerful, flexible, and cost-efficient browser automation backend for their AI agents. It's particularly well-suited for building web scrapers, monitoring tools, or complex workflow automations where reliability, stealth, and low token usage are critical.