Reflex
Peramban yang lebih cepat dan paralel untuk agen AI: ia mengendalikan Chrome Anda sendiri dari Claude Desktop atau klien MCP mana pun, dengan perjalanan bolak-balik dan token yang lebih sedikit dibandingkan Playwright MCP.
Dokumentasi
A faster browser for your AI agent
Research products, fill forms, compare prices, scrape tables. All in plain English, in your own Chrome.
25 free calls the moment you install. No signup, no card.
Reflex MCP is a local MCP server that drives your own Chrome from any MCP client. It returns compressed page snapshots and batched actions instead of full dumps, so your agent finishes the same task up to 2.2x faster end to end on a fraction of the context, and never sends page content to our servers.
Already in Cursor, VS Code, or Claude Code? Add Reflex in one click.
How does Reflex MCP compare to Playwright MCP?
Reflex is the Playwright MCP alternative when every page look costs your context window: our head-to-head measured 6 tool calls and ~13K tokens across three flows versus 14 calls and ~56K for Playwright MCP, with 5 of 6 flows completed versus 3. Playwright MCP is free and better for iframes, network mocking, and Firefox/WebKit; Reflex wins on round trips, tokens, parallel sessions, and forensic failures.
Full Reflex vs Playwright MCP comparison · Benchmarks and methodology
| Playwright MCP | Reflex | |
|---|---|---|
| Tool calls, 3 measured flows | 14 | 6 |
| Tokens into context, same flows | ~56K | ~13K |
| Flows completed, of 6 attempted | 3 | 5 |
| One look at a Wikipedia article | ~31K tokens | ~4K tokens |
| One look at the worst page tested (W3C CSS spec) | ~169K tokens | ~7K tokens |
| Re-reading an unchanged page | ~16K tokens | 91 characters: "(page unchanged)" |
Measured 2026-06-12. See every number and how to reproduce it.
Clients like Claude Desktop pay that snapshot cost on every look. Reflex compresses the page instead, so your agent keeps its context for the real work. Playwright MCP is free and better for iframes and other browsers; Reflex wins when every look costs you context. Full methodology and the reproduce command are on the benchmarks page.
Can I run multiple browsers in parallel?
Yes. Pass a session name on any Reflex tool call and it spins up a separate, isolated Chrome on demand, free. One agent can research ten sites at once; subagents can each drive their own browser. Playwright MCP shares a single browser, so parallel agents collide on the same tab. Reflex is the browser automation MCP built for parallel work from Claude Desktop, Cursor, and other MCP clients.
- flights
- hotels
- cars
one isolated Chrome each, running side by side
reflex_open(url, session: "flights")
Watch the token race.
Wikipedia: search an article, open its revision history
Playwright MCP5/5 calls
tokens into context~30K
Reflex2/2 calls
tokens into context~8K
One real flow above, measured 2026-06-12. Across all three verified flows: 6 calls / ~13K tokens vs 14 calls / ~56K tokens (~2 calls per flow with Reflex).
Two minutes to your first task.
Built for agents, not just browsers.
Failures you can act on
When a step misses, Reflex sends back the closest matches, the console errors, and a screenshot, so your agent fixes it on the next try instead of guessing.
Checks built into the action
Reflex verifies the result inside the same call, so a task clearly succeeds or fails without an extra round trip to confirm.
0-token replay
Save a task and re-run it anytime with no AI and no tokens. It repairs itself when the page changes.
Your pages never leave your machine.
Reflex runs entirely on your computer: the MCP server, the browser, your logins, every page you visit. The only things that ever cross the network are your API key and a count of tool calls, so we can meter credits. We could not read your pages, sessions, or passwords if we wanted to; the architecture has no path for them.
Compare that to cloud browser products, where every page you visit renders on someone else's server.
Reflex servers
- Is this key valid?
- Credit balance
- Usage counts
- Paddle payments
No path exists for your pages, logins, or sessions to reach us.
API key + call countpages never cross
Your machine stays here
Claude Desktopor Cursor, Claude Code, any MCP client
Reflex MCP serverruns locally
Your Chromeyour real browser and logins
Pages, sessions, passwordsread and acted on locally
up to 2.2x
faster end to end on 6 live-site tasks, with a real agent driving (vs Playwright MCP)
7.9x
fewer tokens per page look vs Playwright MCP, averaged across 12 live sites
0
tokens to re-run a saved flow (deterministic replay, no model in the loop)
See every measured number and how to reproduce it
Pay for what you use. Paid credits never expire.
Free
$0
Kick the tires, no card.
- 25 calls with no account at all
- 30 more credits when you sign up
- ~10-15 tasks · free credits expire in 14 days
Starter
$9one-time
For occasional flows.
- 150 credits
- ~50-75 tasks
- $0.12-0.18 per task
Most popular
Pro
$19one-time
Best value for steady use.
- 500 credits
- ~165-250 tasks
- $0.08-0.11 per task
Power
$49one-time
For heavy, daily automation.
- 1,800 credits
- ~600-900 tasks
- $0.05-0.08 per task
Questions, answered straight.
Everything runs locally: the MCP server, your Chrome, your logins, the pages. The only data that reaches our servers is your email (for your account), your API key, and a counter of tool calls for credit metering. Page content, URLs you visit, credentials, and session cookies never leave your machine; there is no code path that could send them. Payments are handled by Paddle, our merchant of record; we never see card numbers. Our privacy policy is short because there is almost nothing to disclose.