aginxbrowser

AginxBrowser

English | 中文

The Browser for AI Agents. See the live web. Read it. Act on it. Remember it.

A browser built for agents from the first line of code — not a human browser bolted onto automation. See the world, read it, search it, act on it, and keep what you read: one Rust binary with built-in V8, no Chromium required.

Humans have Chrome. Agents have AginxBrowser.

One binary, zero dependencies, instant service. The HTTP API is the whole interface — agents plug in and go.

*Historical clip from the v0.5.3 release assets (current product is v0.6.0). The star that got our attention: Pierre Tachoire, co-founder of Lightpanda — the headless browser our bench measures against — starred the repo. 90 seconds on why that mattered to us.*

*Real pages rendered by AginxBrowser's diting engine (no Chromium) — Wikipedia, this repo, Rust. Screenshot it yourself →*

Why Agents Need Their Own Browser

Measured against headless Chrome on the same 20 pages, same network (bench, 2026-08-28): 7.6× faster to agent-usable text (p50 532 ms vs 4 053 ms), ~10× less memory (227 MB for the whole process vs ~2.1 GB per Chrome page), and 0 hard failures where Chrome's --dump-dom produced no DOM on 5 of 40 loads. Re-run on v0.5.21 (2026-09-29) on a degraded-network day held the ratio — 5.4× p50, 468 MB whole-run vs 1.75 GB per page — both raw TSVs are committed (those runs are historical; the current release is v0.6.0). An agent's total cost is browser efficiency × model efficiency — this is the browser half.

Existing "browser automation" was built for humans or for one-shot scraping — not for agents:

· AginxBrowser · Puppeteer/Playwright · Firecrawl · Browser-use Designed for · Agents first · Human debugging · Scraping service · LLM wrapper Dependencies · Single binary, no Chromium · Chromium ~500MB · Docker ~1GB · Chromium Sees (screenshots) · ✅ built-in diting rendering engine · Needs Chromium · ❌ · Needs Chromium Reads · markdown + js_extract + fetch receipts · DIY · markdown · DIY Writes documents · ✅ render_markdown: deterministic HTML + inline-SVG diagrams · ❌ · ❌ · ❌ Finds (search) · ✅ 20 engines, 7 categories, merged · ❌ · ❌ · ❌ Acts · indexed session interaction · DevTools API · ❌ · LLM-driven Remembers · ✅ local fetch/search cache (SQLite FTS5) · ❌ · crawl cache · ❌ Protocol · HTTP · Node API · HTTP · Python TLS fingerprints · ✅ Chrome/Firefox/Safari/Edge · Plugin required · ❌ · ❌ CAPTCHA · ✅ detect + auto-wait + optional 2captcha · DIY · ❌ · ❌ Interactive sessions · ✅ persistent · ✅ · ❌ · ✅

Same-tier engines, not the tools in the table above. Cells are capabilities, not speed.

· AginxBrowser · Obscura · Blitz · Lightpanda What it is · Rust browser, V8, diting CSS paint · Rust headless browser, V8 · HTML/CSS engine (Stylo). Not an agent browser · Zig headless browser, V8 Screenshot · built-in paint, opt-in build · screenshots, screencast, PDF · paints a window · Hermes integration falls back to Chrome for screenshots Public CSS suite · none in CI · not published as WPT · WPT in CI, including SVG · not published as WPT License · Apache-2.0 · Apache-2.0 · Apache-2.0 and MIT · AGPL-3.0

An agent needs five things from a browser: see, read, find, act, remember. One binary covers them all — systemd-friendly, zero dependencies.

Core advantage: no Chromium. AginxBrowser inlines a full browser engine (V8 + Rust HTTP stack + the diting CSS/layout/paint rendering engine, with the Blitz/Stylo/Taffy lineage as its reference implementation). From v0.6.0 the JS event loop lives in the owned diting_runtime crate (deno_core retired). No Puppeteer, no Chrome, no Docker. One Rust binary under systemd is your agent browsing infrastructure.

Two Things Stateless Renderers Can't Do

Most new "agent browsers" are stateless, fingerprint-less one-shot renderers — fine for public pages, dead on arrival against Cloudflare or login flows. AginxBrowser goes the opposite way:

• 🔐 Real TLS fingerprints — stealth mode replicates Chrome / Firefox / Safari / Edge TLS handshakes via BoringSSL (JA3 preset names such as chrome145, firefox133 — not just a UA string), switchable per request. The fleet User-Agent and the default TLS shape are one coherent pair: Chrome/145 over the chrome145 preset — UA and ClientHello agree by construction (deliberately 145: some edges route UA majors non-monotonically and wreq-util presets stop at 149; 145 is the only production-routed major a real preset backs). Cloudflare Turnstile challenges wait automatically for cf_clearance. Fingerprint-less engines eat 403s — we get through.
• Evidence: public Stealth/TLS comparison (JA4 / Akamai H2 vs curl & Chrome, Kitesurf/Lightpanda cited not-run) — docs/stealth-tls-comparison.md · 中文 · receipts bench/stealth-tls-20261019/
• Scoreboard: tip micro-bench (cold start, /fetch shortset, session snapshot) vs curl/Chrome; Lightpanda/Obscura not run — docs/scoreboard.md · 中文 · receipts bench/scoreboard-20261019/
• 🤝 Stateful interactive sessions — login state injectable and exportable (session_create(cookies=...) ↔ session_cookies), surviving pagination and multi-step flows; persistent: true even survives idle eviction and server restarts — the same session id comes back logged in. One-shot engines throw state away.

Reference point: Cloudflare's Kitesurf explicitly ships neither real TLS-fingerprint negotiation nor persistent auth sessions — anti-bot and login territory is exactly where AginxBrowser plays.

Apache-2.0 open source, single binary — self-host today, no cloud lock-in.

Every Fetch Is a Receipt

Agents act on what a browser tells them, so the response reports what actually happened — not just "got a 200":

• `tier` — which path served the page: plain HTTP (~100 ms) or the V8-rendered browser tier. An agent can see *why* a fetch was fast or slow.
• `redirected_from` — the full redirect trail. redirected_from[0] is the URL you asked for, url is where the content actually came from — requested paired with effective, every hop visible.
• `content_hash` + `changed_since_prev` — every fetch is hashed; consecutive samples of the same URL can be diffed. A rate-limited origin serving the same frozen 200 body for days reads as changed_since_prev: false — the cheapest drift detector there is.
• `captcha_event` — when a challenge page was detected (and solved, if a solver is configured), the response says so instead of handing over a challenge page as if it were content.

The local cache builds on the same idea: search hits come back with [§ heading] section prefixes so an agent knows *where on the page* a hit landed, and ranking fuses keyword relevance with freshness.

Capabilities

• Tiered rendering: static pages over plain HTTP (~100ms); V8 spins up only when JS rendering is needed (~1-2s) — 90% of the bench page set served without spinning up V8 at all; every response reports which tier served it (tier field)
• Multi-engine meta-search: general web (Baidu / Bing / Sogou / WeChat / DuckDuckGo / Wikipedia / Hacker News), news (Bing News), code (Stack Overflow, GitHub, MDN), packages (npm, PyPI, RubyGems), academic (arXiv, OpenAlex), AI models (Hugging Face) — 20 engines across 7 categories, queried concurrently, merged and deduplicated. Operators can plug a private Meilisearch index into the same /search. Search → read in one step
• Image search: categories=images hits Baidu/Bing image indexes and returns direct binary image_url links (downloadable straight to jpg/png) plus source_url provenance
• Interactive sessions: persistent browser sessions with indexed interaction (state/click/input/scroll/eval) — agents browse like humans do, and session_export turns what an agent figured out into a runnable curl replay script (zero model tokens on re-run) — or, with format=json, into a flow document (flow_run replays it server-side with {{var}} substitution, wait/expect gates and saved outputs; installed flows live in workflow//flow.json, dropped in without a rebuild). Session tools also cover the acting part: session_viewport simulates device viewports (media queries respond), session_wait blocks on a selector or predicate with a timeout, session_screenshot renders the live state, session_console replays the page's console ring, and session_storage exports/restores cookies plus localStorage for login hand-off
• Playback-link sniffer: session_network(filter=media) extracts the m3u8/mp4/dash URLs a page's player *actually requested* at runtime — links found only in page HTML are often decoys, so the request log is the source of truth. GET /session/{id}/har exports the same traffic as HAR 1.2 (retained bodies included)
• File download: streaming to disk (no memory buffering), SHA-256 integrity, resume of interrupted transfers — for binaries, archives, datasets
• Local cache that remembers: every fetch/search lands in SQLite (FTS5) at ~/.aginxbrowser/cache.db — a re-fetch inside the TTL answers from what the agent already read instead of re-paying network time, with CJK substring matching, [§ heading] section-aware snippets, per-URL content hashes for drift detection, and per-session scoping for shared deployments
• CAPTCHA handling: type detection with automatic Cloudflare challenge wait and optional 2captcha integration — search never stalls on verification pages
• JS data extraction: js_extract pulls window.__INITIAL_STATE__ and other structured data out of SPAs
• Document generation: POST /render_markdown turns markdown into a deterministic, self-contained HTML artifact — the document layer, so agents never write HTML by hand. Prose rides a plain offline shell (no fonts, no scripts); fenced archify blocks carry typed zero-coordinate diagram JSON (sequence / workflow / architecture / dataflow / lifecycle families) and render to inline SVG via the layout engine. Same input, same bytes — the receipt carries the sha256 so determinism is verifiable. theme (light/dark) and preset (classic / signal-flow / blueprint / editorial) bake colors at generation time; quality: "showcase" is the delivery gate, grading route crossings, label clearance and rhythm without touching the artifact bytes. Guided-view tabs plus window.agxViewer (focus / ego / route / reach) make the artifact interactive. Mermaid sources are the agent's job to translate into archify JSON, not the engine's. Diagram vocabulary adapted from archify (MIT)
• Screenshot rendering: /screenshot endpoint (opt-in --features screenshot) paints the JS-rendered DOM with the diting rendering engine — pure CPU, no Chromium — to PNG. Vision input for agents
• Timeline video: /video renders a page's animation timelines to MP4 — the page's scripts register GSAP-style timelines in window.__timelines (duration() + pause(t)), each frame seeks to t=i/fps and paints the viewport, and the frames pipe into ffmpeg (H.264, yuv420p). Deterministic by construction: no wall clock in the pixel values, same render twice = same MP4. Needs ffmpeg on PATH
• Page set (PDF/PNG/PPTX/DOCX): /pdf cuts a rendered page into pages and packages them — print mode paginates at top-level block boundaries (default A4 @96dpi, no half-cut text where a break can land on a block edge), slides mode makes one page per CSS-selector match sized to the element (an HTML deck with one .slide per page exports as a real deck). Image-based PDF: per-page JPEG via DCTDecode, hand-rolled PDF 1.4 writer, zero new dependencies. PPTX packages the same pages as one slide per page; DOCX as one page-sized section per page — both hand-rolled OOXML (stored-ZIP writer, fixed timestamps), byte-deterministic, zero new dependencies
• TLS fingerprint spoofing: stealth mode impersonates Chrome/Firefox/Safari/Edge TLS shapes (default preset chrome145; fleet UA Chrome/145), switchable per request
• Firecrawl compatible: /v1/scrape endpoint — existing Firecrawl clients migrate by changing the base URL
• DNS rebinding protection: built-in SSRF guard + post-resolution IP validation

A Browser, Not a Crawler

AginxBrowser exists for real-time retrieval: an agent arrives with a question, reads a handful of pages, leaves with the answer. It is not a crawling tool — and the product is shaped so it can't quietly become one:

• robots.txt is not our gate. The RFC 9309 checker ships built in, but a real-time lookup layer isn't a crawler and doesn't do crawler etiquette by default; operators who want it set AGINXBROWSER_HONOR_ROBOTS=1.
• No site-walking API. There is no crawl endpoint and no link-following recursion — every page load happens because an agent asked for that page.
• Built-in budgets. Per-domain: 20 pages/minute. Per interactive session: 200 pages. Toggled via AGINXBROWSER_DOMAIN_RATE_PER_MIN / AGINXBROWSER_SESSION_PAGE_LIMIT (0 disables on your own instance). Generous for an agent grinding through docs or a console; fatal to the page-after-page crawl pattern, including subdomain rotation (one registrable domain, one budget).
• The hosted instance (browser.aginx.net) runs tighter budgets. Every user shares one egress IP, and keeping sites comfortable with that IP is part of the service. Self-host if you want different numbers.
• Need to bulk-crawl a site? Use a crawler. This isn't one, and it won't become one.

What It's For

Not demos — real jobs agent browsers are doing today:

• Grind through admin consoles — AWS / App Store Connect / Google Play, dozens of menu layers per task. Let the agent click; it comes back only when authorization is needed.
• Batch actions behind login — fill carts, dig through order history, check pages that only render while logged in. Inject cookies, operate, export for reuse.
• Past anti-bot walls — Cloudflare protection, Turnstile challenges, TLS fingerprint checks. Stealth mode pushes through instead of retreating at 403.
• The Chinese internet — Baidu / Sogou / WeChat meta-search across 5 engines, correct Chinese page rendering. Not English-web-only.
• On-the-spot scripting — agent reads the page, writes JS, evals it: highlighted comparison tables, reflowed content, product filters on hidden parameters. GreaseMonkey-on-steroids.
• Multimodal vision — screenshots as visual input for look-and-judge flows: picking seats, recognizing layouts, verifying rendering.

Where It Sits in the Computer-Use Stack

Computer-use agents come in two layers. GUI-layer stacks (Cua, desktop CUA agents) drive a whole machine: screenshots of a display in, X11 mouse/keyboard events out, a VM or container per session. Engine-layer browsers skip the desktop — the page itself is the machine. AginxBrowser is the engine layer:

· GUI layer (desktop CUA) · AginxBrowser (engine layer) Action space · screen pixels → OS input events · DOM: click by coordinates or selector, real event dispatch State readout · screenshot of a display · structured DOM / accessibility tree, screenshot optional Per-session cost · Xvfb + VM/container · one process, no display server Frames · flattened into pixels · first-class: hit-testing descends into iframes, frame-scoped evaluate Replay · best-effort (compositor timing) · action logs export as runnable curl / flow JSON

Quick Start

One-command full install (SKILL.md trigger surface + verification):

# Download -> inspect the contents -> run only after review (never blind-run network scripts)
curl -fsSL https://raw.githubusercontent.com/yinnho/aginxbrowser/main/skill.sh -o skill.sh
less skill.sh
bash skill.sh

Install the skill trigger surface via skills.sh:

npx skills add yinnho/aginxbrowser

Or just hand this block to your AI agent — it sets itself up:

Set up aginxbrowser as your web access layer:
1. Install v0.6.0 locally and verify it is healthy:
   # Preferred — current release binary (install.sh picks the platform):
   curl -fsSL https://browser.aginx.net/install.sh -o install.sh && less install.sh && bash install.sh
   # Optional Homebrew (tap tracks v0.6.0; still verify with `aginxbrowser --version`):
   #   brew trust yinnho/aginxbrowser && brew install yinnho/aginxbrowser/aginxbrowser
   aginxbrowser doctor   # then start: aginxbrowser (listens on 0.0.0.0:8089)
   curl http://127.0.0.1:8089/health
2. From now on, route web tasks through http://127.0.0.1:8089 instead of raw curl:
   - read a page / scrape a URL  -> POST /fetch {"url":...}        (markdown output, Cloudflare bypassed by default)
   - search the web              -> POST /search {"q":...,"fetch_top":3}
   - see a page                  -> POST /screenshot {"url":...}
   - login / form / click-through-> POST /session/create -> /session/{id}/state -> /input /click -> /close

Self-hosting:

# Prebuilt binary for current release v0.6.0 (platform detect + sha256 + mirror fallback)
# macOS / Linux / Windows (git-bash). Assets: https://github.com/yinnho/aginxbrowser/releases/tag/v0.6.0
# Cautious: download -> inspect -> run (never blind-run network scripts)
curl -fsSL https://browser.aginx.net/install.sh -o install.sh
less install.sh && bash install.sh
# Or straight in, if you trust the repo:
#   curl -fsSL https://browser.aginx.net/install.sh | sh
# GitHub slow/blocked? AGINXBROWSER_GH_PROXY=https://ghfast.top/ bash install.sh
aginxbrowser doctor   # features + fonts + egress self-check

# Optional: Homebrew (7+ needs the trust line first). Tap is at v0.6.0 —
# still verify with `aginxbrowser --version` after install / upgrade.
# brew trust yinnho/aginxbrowser
# brew install yinnho/aginxbrowser/aginxbrowser
# brew upgrade aginxbrowser

# Or build from source (--features stealth,screenshot or you lose both)
cargo build --release --features stealth,screenshot

# Start the service
./target/release/aginxbrowser
# → Listening on 0.0.0.0:8089

# Verify
curl http://127.0.0.1:8089/health
# → {"status":"ok","engine":"diting"}

# Fetch a page
curl -sS -X POST http://127.0.0.1:8089/fetch \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com"}'

# Search (snippets only by default; fetch_top grabs page bodies for the top N)
curl -sS -X POST http://127.0.0.1:8089/search \
  -H "Content-Type: application/json" \
  -d '{"q":"macbook price","max_results":5,"fetch_top":2,"max_chars_per":2000}'

# Create an interactive session
curl -sS -X POST http://127.0.0.1:8089/session/create \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com"}'
# → {"session_id":"s_1","url":"https://example.com/"}

REST Routes

Every capability is plain HTTP — no SDK required. There is no /openapi.json (the routes are few enough to list here); full request/response fields for each route are in docs/API.md.

Method · Path · What it does GET · /health · Liveness + build commit, UA, TLS, capabilities GET · /doctor · Deep self-check: engines live, fonts, egress GET · /engines · Search engine catalog with live suspension state POST · /fetch · Fetch a page → text/markdown/html. Params: render_tier (auto/http/browser), max_chars, capture_xhr, … POST · /search · Multi-engine meta-search. Params: engines, categories, max_results, time_range (day/week/month/year), fetch_top, max_chars_per, wait_secs POST · /click · Click a CSS selector on a page POST · /eval · Evaluate JavaScript on a page POST · /download · Stream a file to disk (sha256, resume) POST · /screenshot · Render page → PNG (screenshot feature) POST · /video · Render animation timelines → MP4 (screenshot feature) POST · /pdf · Paginate page → PDF/PNG/PPTX/DOCX (screenshot feature) POST · /v1/scrape · Firecrawl-compatible scrape (+ actions) POST · /flow/run · Run a recorded/edited flow JSON to completion — zero model tokens (name runs an installed workflow//flow.json, flow is inline, vars substitute {{placeholders}}, session_id composes with imported login state) POST · /session/create · Start an interactive session (cookies/UA carried across calls) POST · /import/curl · Create a logged-in session from a DevTools "Copy as cURL" GET · /session/list · Live sessions POST · /session/{id}/navigate · /state · /click · /click_xy · /drag · /input · /scroll · /eval · /wait · /dialog · /viewport · /screenshot · /clone · /close · Session actions GET · /session/{id}/cookies · /storage · /console · /network · /har · /export · Session inspection POST · /render_markdown · Markdown → deterministic self-contained HTML artifact (+ inline-SVG diagrams)

Every capability is one POST away — no SDK, no protocol adapter.

Project Layout

aginxbrowser/
├── ARCHITECTURE.md       # Layering legislation (authoritative)
├── Cargo.toml            # Workspace root (members = crates/diting{,_ops,_runtime})
├── crates/
│   ├── diting/           # Engine crate: DOM/CSS/layout/paint/JS/net/Page
│   │   ├── build.rs      # V8 snapshot generation
│   │   ├── js/           # bootstrap.js + companion scripts
│   │   └── src/          # diting_{dom,css,net,js,layout,fonts,browser}
│   ├── diting_ops/       # #[diting_op] glue generator (engine companion)
│   └── diting_runtime/   # Owned JS event loop / op settlement
├── README.md
├── docs/
│   └── API.md            # Full API reference (HTTP)
├── bench/                # Benchmark harness + results (vs headless Chrome)
│   ├── README.md         #   methodology + numbers
│   ├── pages.txt         #   fixed 20-page set
│   ├── run.py            #   harness
│   ├── summarize.py      #   TSV → results table
│   └── results/          #   raw run data
└── src/
    ├── main.rs              # HTTP/CLI assembly (thin face)
    ├── routers/             # HTTP handlers: acquisition/sessions/outputs/render
    ├── server/              # Acquire façade by job: browser/fetch/outputs/search
    ├── session/             # Interactive sessions (manager/interact/state/…)
    ├── flow/                # Flow interpreter + install
    ├── search/              # Metasearch orchestration + per-engine adapters
    ├── store/               # Local fetch/search cache (SQLite FTS5)
    ├── docgen/              # Document layer: markdown → HTML + SVG diagrams
    ├── screenshot.rs        # Eye-side HTML→PNG (keep monolithic)
    ├── render.rs            # Tiered rendering (HT