# Ostler > Ostler is an open-source MCP browser for AI coding agents such as Claude Code, Codex and Cursor. It reads, researches and acts in a real Chrome, and checks its own work: research answers cite only pages it actually read, with every quote verified; forms are filled in one call with every value read back; and Qdrant remembers what worked. Ostler is one local Rust binary (Apache-2.0) that runs Chrome and speaks MCP over stdio to coding agents and A2A over HTTP to other agents. It has no LLM inside, no cloud browser and no anti-bot evasion. Status: v0.2, early (development source; no tagged release yet); tested against real Chrome and Qdrant on local fixtures, not yet benchmarked on the open web. Key facts: - Install: `pip install ostler-mcp` (or `uvx --from ostler-mcp ostler serve`), then `ostler start` (connects every AI app it finds) or `ostler setup claude --apply` (Claude Code); or see the agents page for Codex, Cursor and VS Code. - Other ways to install (v0.2.0): the install script for macOS and Linux (https://inamdarmihir.github.io/ostler/install.sh), Claude Desktop `.mcpb` (macOS), the VS Code extension on Open VSX (https://open-vsx.org/extension/inamdarmihir/ostler), and `cargo install --git https://github.com/inamdarmihir/ostler ostler` from source. Windows (x86_64, experimental) and a Docker image (`ghcr.io/inamdarmihir/ostler`) arrive in v0.2.1. Not published yet: the VS Code Marketplace, Homebrew, npm and crates.io. - Requirements: Chrome or Chromium; Python 3.8+ for pip (no Rust); Rust 1.89+ only for the from-source install. - Evidence: local fixtures; a five-task scripted comparison with Playwright MCP and Chrome DevTools MCP; seven live public practice pages. Not yet measured on the open web. - Browsing: a general MCP browser in a real Chrome. 16 core tools: navigate, snapshot, read, click, type, tabs, screenshots and `look` (numbered click targets). Log in once with `--profile-dir`; CAPTCHAs are handed to a person, never solved. - Research: `read` returns the best passages for a query; `cite` marks a quote VERIFIED only if its words are on a page read this session; `evidence` lists citations and the pages that could not be read (paywall, bot wall, robots.txt). - Forms: `form_fill` fills a whole form in one call and reads every value back; 3 tool calls instead of 22 on a 20-field form (same engine, one call per field; scripted, model-free, local fixtures). - Memory: optional, in Qdrant or an encrypted local file; stores facts you save and per-site playbooks (structure, not values). Secret-looking values are refused; detection is heuristic. Tool output goes to your MCP client and, through it, to that client's model provider. - Safety: page text is fenced as untrusted, a domain allowlist holds every request, CAPTCHAs are handed to a person. ## Docs - [Get started without programming](https://inamdarmihir.github.io/ostler/docs/get-started.html): Install Ostler with a double-click, an editor extension, pip or uvx, or one command: no Rust and no config files. - [Quickstart](https://inamdarmihir.github.io/ostler/docs/quickstart.html): Install, check, connect, then read and cite a demo page or fill a demo form, in five minutes. - [Install](https://inamdarmihir.github.io/ostler/docs/install.html): Every way to get the binary, what works today, update and uninstall. - [Use with your agent](https://inamdarmihir.github.io/ostler/docs/agents.html): Add Ostler as an MCP browser to Claude Code, Codex, Cursor, VS Code or any MCP client, with a forms and a research workflow. - [LangChain, LlamaIndex, AutoGen](https://inamdarmihir.github.io/ostler/docs/frameworks.html): Use Ostler's tools from three Python agent frameworks: tested code, real output, and the one-session rule for cite. - [Research with checked citations](https://inamdarmihir.github.io/ostler/docs/research-with-citations.html): Read pages, check each quote against them, and report what could not be read, with real tool output. - [Cookbook](https://inamdarmihir.github.io/ostler/docs/cookbook.html): Recipes: deep research with checked citations, reading, QA, logins, forms, approvals and A2A, with what is not built yet. - [Forms and verification](https://inamdarmihir.github.io/ostler/docs/forms.html): How keys become fields, how values are verified by read-back, and what submit results mean. - [Memory and privacy](https://inamdarmihir.github.io/ostler/docs/memory.html): What is stored, what sealing hides and does not, local embeddings, per-tenant tokens. - [Configuration and CLI](https://inamdarmihir.github.io/ostler/docs/configuration.html): Config file, environment variables, every serve and a2a flag. - [Tools reference](https://inamdarmihir.github.io/ostler/docs/tools.html): All 32 MCP tools: browse, read and cite, search, extract, forms and memory, with parameters and profiles. - [Security model](https://inamdarmihir.github.io/ostler/docs/security.html): The threat model: what Ostler does, and what it does not do. - [A2A](https://inamdarmihir.github.io/ostler/docs/a2a.html): Let other agents hand Ostler forms over Agent2Agent v1.0. - [Architecture](https://inamdarmihir.github.io/ostler/docs/architecture.html): Crates, the injected runtime, the form model, and the decisions behind them. - [Benchmarks](https://inamdarmihir.github.io/ostler/docs/benchmarks.html): Measured on local fixtures with a scripted agent. Read the caveats. - [Ostler vs Playwright MCP](https://inamdarmihir.github.io/ostler/docs/ostler-vs-playwright-mcp.html): Five scripted form tasks on local fixtures: tool calls, tokens, time, and what the comparison does not show. - [Economics](https://inamdarmihir.github.io/ostler/docs/economics.html): Unit costs, scaling limits, the moat question and what could go wrong. - [Design review](https://inamdarmihir.github.io/ostler/docs/review.html): The October 2026 review: five design flaws and their status. - [Troubleshooting](https://inamdarmihir.github.io/ostler/docs/troubleshooting.html): Symptoms, causes and fixes. Start with ostler doctor. - [Changelog](https://inamdarmihir.github.io/ostler/docs/changelog.html): What changed, including the Unreleased section. ## Optional - [Full documentation as one Markdown file](https://inamdarmihir.github.io/ostler/llms-full.txt) - [Source code on GitHub](https://github.com/inamdarmihir/ostler) - [README](https://github.com/inamdarmihir/ostler/blob/main/README.md)