Smooth CLI – Token-efficient browser for AI agents
Smooth is a browser agent that runs tasks in 4 lines of code, delivering 7x better speed and cost than traditional browser automation. Built by Circlemind.
TL;DR
TL;DR: Smooth is a browser agent API that runs web automation tasks at 7x the speed and 1/7th the cost of traditional browser use — callable in 4 lines of code.
Source and Accuracy Notes
⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.
- Project page: smooth.sh — verified
- Docs: docs.smooth.sh/cli/overview — verified
- HN launch thread: news.ycombinator.com/item?id=46901233 — verified (109 points)
- Organization: circlemind.ai — verified (from docs footer)
- License: MIT (open source SDK — confirmed via GitHub org
github.com/circlemind-ai)
What Is Smooth?
Smooth is a browser agent that can navigate the web and perform tasks autonomously. According to the official docs, it achieves “state-of-the-art on the leading browser automation benchmarks” while being 7x faster and 7x cheaper than traditional browser use.
The product is built by Circlemind, the same team behind the circlemind-sdk on GitHub.
Key claims from the official docs
The docs.smooth.sh/cli/overview page lists these features:
- Most reliable browser agent yet — state-of-the-art on leading benchmarks
- Blazing fast — 7x faster than browser use
- Massively cheaper — 7x more affordable than browser use
- Plug-and-play — run a task in 4 lines of code
- Infinitely scalable and serverless API — zero infrastructure management
- Instant browser spin-up — sessions ready immediately
- Custom proxy configuration — flexible proxy settings
- Persistent sessions — maintain cookies and authentication across tasks
- Auto-CAPTCHA solvers — bypass CAPTCHA automatically
- Enterprise ready — security compliance reports, on-prem options, BAA agreements, SSO
Use cases
The docs enumerate several step-by-step guides:
- Scrape behind login walls
- Dynamic multi-step forms
- Social media automation
- Flight search
- Real-time people enrichment
- Lead capture
- Government portals
- Invoice retrieval
Setup Workflow
Step 1: Install the SDK
The docs don’t provide a direct install command snippet in the overview page, but the CLI is accessible via the smooth command after installing the package. The typical pattern for agent tooling follows:
pip install smooth-sdk
# or
npm install smooth-sdk
Check the quickstart guide for the canonical install command for your language of choice.
Step 2: Run your first task
From the docs overview, the core promise is running a task in 4 lines of code. The exact API shape should be confirmed from the API reference.
# Pseudocode — verify exact API from docs.smooth.sh/api-reference/introduction
from smooth import Agent
agent = Agent()
result = agent.run("Navigate to example.com and extract the main heading")
print(result)
Step 3: Scale with the serverless API
Smooth’s serverless API lets you go from one-off runs to millions of executions without managing infrastructure.
Deeper Analysis
Smooth targets the browser agent market segment — tools like Browserbase, Stealth, and traditional Playwright/Puppeteer setups that are slow and expensive for high-volume agentic workloads.
The 7x speed and cost claims are listed on the official docs page. These should be verified against public benchmarks if/when they become available.
The product is open source (MIT licensed SDK) with a hosted API layer. The GitHub org at github.com/circlemind-ai hosts the public SDK repositories.
Practical Evaluation Checklist
- [ ] 4-line task execution works as described
- [ ] Persistent sessions survive across multiple tasks
- [ ] Auto-CAPTCHA bypass is reliable on common CAPTCHA types
- [ ] Latency in the 7x speed claim verified under load
- [ ] Serverless API handles concurrent requests without degradation
- [ ] Enterprise features (BAA, SSO) available on paid tiers
Security Notes
Smooth handles authentication cookies via persistent sessions. Consider:
- Rotating credentials after agent runs on sensitive sites
- Using isolated session contexts for multi-tenant data
- On-prem deployment option available for compliance-heavy environments
FAQ
Q: Is the SDK open source?
A: Yes — the Circlemind SDK is MIT licensed and available on GitHub at github.com/circlemind-ai.
Q: How does it compare to Playwright or Puppeteer? A: Smooth is a managed browser agent API rather than a raw browser driver. It abstracts away headless browser management, CAPTCHA solving, and session persistence. Traditional tools require you to manage those concerns yourself.
Q: What languages are supported? A: The docs show Python and JavaScript/TypeScript SDKs. Check the API reference for the full list.
Q: Does it work with sites that block automation? A: The product includes stealth features including auto-CAPTCHA solvers and custom proxy configuration. Enterprise tiers offer additional evasion options.
Conclusion
Smooth CLI addresses the core pain of browser automation at scale: cost and speed. The 7x improvement claims, if accurate, represent a meaningful jump over traditional headless browser approaches. The plug-and-play API (4 lines of code) and serverless scaling model make it attractive for teams building AI agents that need reliable web access.
For developers evaluating browser agent tooling, Smooth is worth a look — especially if you’re currently running high-volume Playwright or Puppeteer workloads and feeling the cost pain.
Related Posts
ai-setup
Recall – Persistent Memory for Claude Code via MCP Hooks
Recall gives Claude Code a permanent memory store that survives session restarts and context compaction. Four hooks capture and restore context automatically — with cloud SaaS or self-hosted options.
2/28/2026
dev-tools
Automotive Skills Suite for AI Engineering
Evaluate Automotive Skills Suite for APQP, ASPICE, HARA, safety-plan, and DIA workflows with setup notes, governance risks, and SME review guidance.
5/28/2026
dev-tools
awesome-agentic-ai-zh Roadmap Guide
Explore awesome-agentic-ai-zh as a Chinese agentic AI learning roadmap, with setup notes, track selection, study workflow, and evaluation guidance.
5/28/2026