Leapfrog

Multi-session browser MCP for AI agents.
36 tools. 15 parallel sessions. Stealth. HUD. Self-improvement. Up to 10x fewer tokens.

npm i leapfrog-mcp| Works with Claude Code, Cursor, Windsurf

Playwright MCP sends~14,000 tokensfor a content-heavy page like Hacker News. Most of that is noise. Your context window fills up. Your agent gets confused. You pay for it.

Leapfrog sends~1,400 tokens. Same page. Same information. Up to 10x less noise.

┌─────────────────────────────────────────────────────┐ │ Playwright MCP │ │ ████████████████████████████████████████ ~14,000 │ │ │ │ Leapfrog │ │ █████ ~1,400 │ └─────────────────────────────────────────────────────┘ tokens per page (Hacker News, real test)

Savings range from 2-10x depending on page complexity. Content-heavy pages see the biggest wins. Dense forms see the smallest. The median across real-world sites is~4-5x.

npx leapfrog-mcp --doctor # verify everything works npx leapfrog-mcp --stealth-audit # test all 19 stealth patches npx leapfrog-mcp --config # print MCP config to paste

Add to~/.mcp.json(Claude Code) or your editor's MCP config:

{ "leapfrog": { "command": "npx", "args": ["-y", "leapfrog-mcp"], "env": { "LEAP_MAX_SESSIONS": "15", "LEAP_TILE": "true", "LEAP_HUD": "true", "LEAP_AUTO_CONSENT": "true" } } }

Leapfrog usesplaywright-core(15MB) instead ofplaywright(1.6GB) and doesnotbundle a browser. Either:

- SetLEAP_CHANNEL=chrometo use your installed Chrome/Chromium (recommended)
- Or runnpx playwright-core install chromiumto install the bundled Chromium binary
- Or setLEAP_CDP_ENDPOINTto connect to an already-running Chrome instance

Leapfrog ships 19 anti-detection patches enabled by default (LEAP_STEALTH=true). Four modes:

- true(default) — all 19 patches active
- passive— removes automation signals only (webdriver, HeadlessChrome). Does NOT fake identity (WebGL, fonts, audio). Better for sites where trust matters more than evasion.
- auto— per-domain EXP3 bandit selects the optimal stealth configuration based on what's worked before
- false— no stealth patches

These cover the vectors that fingerprint services like CreepJS and fingerprint-pro actually check:

- Client Hints brands (strips HeadlessChrome)
- navigator.webdriverforced toundefined
- WebGL vendor/renderer (replaces SwiftShader with real GPU strings)
- Connection RTT (non-zero)
- Alert dismiss timing (human-speed delay)
- Window outer/inner height offset
- MIME type array population
- Platform inference from user agent
- chrome.appemulation
- iframecontentWindowprotection
- Media codec spoofing (canPlayType)
- document.hasFocus()override
- Source URL comment stripping
- Custom UA + stealth coexistence (custom user agents no longer disable stealth context)
- CDPRuntime.enabledetection (Error.prepareStackTracefilter)
- Permissions API spoofing (20+ permission types)
- AudioContext fingerprint noise (getChannelData/getFloatFrequencyData)
- WebRTC IP leak prevention (ICE candidate filtering)
- Font enumeration fingerprint spoofing

Per-session stealth control: passstealth: falseinsession_createto disable for a specific session.

SetLEAP_HUMANIZE=trueto enable human-like browser interaction. This is opt-in and adds latency in exchange for more realistic behavior. Six modules:

- Mouse— Bezier curve paths with Fitts's Law timing and micro-tremor jitter
- Typing— Log-normal inter-key delays (200ms median), key dwell time, bigram-aware speed, rollover typing
- Scroll— Inertial simulation with ramp-up and momentum decay (touchpad/mouse-wheel physics)
- Pause— Inter-action "think" delays that simulate cognitive gaps between actions
- Fingerprint— Coherent browser fingerprint generation (platform, device memory, GPU, timezone)
- Utils— Shared math primitives (Box-Muller gaussian, distributions)

Everynavigateandsnapshotcall automatically classifies the page type using weighted signal scoring (no LLM required). 18 types:

login·search-results·product·product-list·checkout·article·dashboard·form·error·challenge·landing·documentation·profile·media·feed·qa·ecommerce·unknown

Classification drives smarter snapshot extraction — login pages surface form fields, articles surface content, dashboards surface interactive elements.

The harness tracks every action in a session and classifies outcomes:

- Action outcome classification—SUCCESS,SILENT_CLICK,NAVIGATION,WRONG_ELEMENT,BLOCKED,ERROR,PENDING
- Bot redirect detection— detects when a site redirects to a challenge or block page after an action
- Loop detection— warns when the agent is stuck clicking the same element, ping-ponging between URLs, or repeating actions
- Session memory—session_memorytool recalls actions after context window compression

Navigate automatically retries with fallback strategies when pages fail to load:
- Tryload(fastest) — if empty, retry withnetworkidle(10s cap)
- Ifnetworkidletimes out (Amazon, ad-heavy sites), fall back todomcontentloaded
- If blocked/challenged, escalate stealth: random delays → wait for JS challenge → rotate session with fresh fingerprint
- Profile sessions (auth'd) never have their session destroyed — hard-capped at Level 2

Opt-out withautoRetry: falseonnavigate. Control max escalation withmaxRetryLevel(0-5, default 3).

Export a session's action history as a replayable recording, then replay it in new sessions:

- session_export— creates parameterized JSON or Playwright script from session history.@eNrefs resolved to stable CSS selectors. Auto-detects emails, passwords, URLs as{{placeholders}}.
- session_replay— replays a recording with parameter overrides. SupportsonError: 'stop'or'skip'.

Turn one-off agent workflows into reusable automations.

Extract data across multiple pages in a single tool call:

- Click-next— auto-detects "Next" buttons, pagination links, "Load more" buttons
- Infinite scroll— scrolls and waits for new content via DOM hash comparison
- URL pattern— increments?page={page}or custom patterns

Replaces 3-4 tool calls per page. Cap: 50 pages, 100K total chars. Stops on: no next button, empty page, duplicate content, or bot detection.

Thedifftool returns only what changed since the last snapshot — additions, removals, changes. Massive token savings for monitoring and polling workflows.

When running headed, Leapfrog overlays visual feedback on every session:

- Click ripple— expanding green circle at click coordinates (agent actions only)
- Zoom-to-target— browser zooms to 1.15x on the clicked element briefly so agents (and humans) can visually track what's happening in tiled windows
- Scroll-to-target— scrollIntoView before clicks so you can see what the agent is about to click

Minimal by design. No borders, no status bars, no cursor overlay — just the feedback that matters.

Multi-Terminal Tiling (LEAP_TILE=true)

Multiple Leapfrog instances share the screen via file-based coordination. Each instance tracks its own windows and a TilesCoordinator assigns global grid slots — no overlap, no manual arrangement. SetLEAP_TILE=true(orLEAP_TILE=masterfor the primary instance). Padding between tiles is configurable withLEAP_TILE_PADDING(default 8px).

Leapfrog auto-detects situations that need a human — CAPTCHAs, login forms, OAuth redirects, Cloudflare challenges — and tries to self-resolve before pausing.

- Auto-resolves first:clicks reCAPTCHA checkboxes, Cloudflare verify buttons, generic verify/continue buttons, then a second-pass retry — all before asking for help
- External solvers:setLEAP_CAPTCHA_PROVIDER+LEAP_CAPTCHA_API_KEYfor CapSolver, 2Captcha, or NopeCHA integration
- Learns what works:remembers which resolution method succeeded per domain and tries the known-good method first on revisit
- Detects reCAPTCHA, hCaptcha, Turnstile, login forms, OAuth redirects, Cloudflare challenges
- Tab title changes to "NEEDS HUMAN" when intervention is needed
- wait_for_humantool — agent calls when stuck, blocks until you resolve it or navigate past

Per-session Playwright tracing with screenshots + DOM snapshots. Export ZIP files viewable attrace.playwright.devvia thesession_export_tracetool. Auto-saves on session destroy.

Leapfrog learns from every visit. Per-domain knowledge persists at~/.leapfrog/domains/{domain}.json— 9 dimensions, all automatic:

LRU eviction at 500 domains. Inspect with thedomain_knowledgetool.

URL validation blocks hex-encoded IPs (0x7f000001), octal notation (0177.0.0.1), CGNAT ranges (100.64.0.0/10), and redirect chains that resolve to internal addresses. Localhost and127.0.0.0/8are allowed by default for local dev workflows — setLEAP_BLOCK_LOCALHOST=trueto block them.

Leapfrog uses pond metaphors to keep things memorable. Your agent is the frog.

- Node.js >= 20
- Chromium — use system Chrome (LEAP_CHANNEL=chrome) or install vianpx playwright-core install chromium

MCP server for browser automation via Vercel agent-browser CLI

Convert any URL to LLM-ready Markdown via real Chrome browsers. 3 tools: scrape, crawl, search. Free via MCP, pay-per-use via x402.

An MCP server that allows AI agents to control a web browser using the browser-use library.

Exposes Chrome browser functionality to AI assistants for automation, content analysis, and semantic search via a Chrome extension.

Control a Chrome browser through the Chrome DevTools Protocol (CDP) from MCP clients for browsing, inspection, and automation workflows.

Stealthy, LLM-drivable browser engine — a Python library + 85-tool MCP server on stealth Chromium with full Chrome DevTools Protocol.

A Go-based MCP server for interacting with the Lightpanda Browser using the Chrome DevTools Protocol (CDP).

A browser automation agent using the Model Context Protocol (MCP) to enable browser interactions.

A browser automation service for capturing console output, useful for tasks like public sentiment analysis.

Control the Chrome browser for web automation using an AI model. Requires the MCP Chrome extension.