Steel
Also known as: Steel Browser, Steel.dev
Open source browser API for AI agents that runs cloud or self hosted Chromium sessions with anti detection, session persistence, and a live session viewer.
Steel is an open source browser API purpose built for AI agents, a batteries included browser sandbox that lets an agent automate the web without managing infrastructure. The tagline captures the idea: humans use Chrome, agents use Steel. It wraps Chromium behind a REST and WebSocket API, handling session lifecycle, state persistence, and anti detection out of the box, and exposes full Chrome DevTools Protocol access so existing Puppeteer, Playwright, and Selenium scripts connect directly. The project is open source with a Docker container for local or self hosted deployment, and a managed cloud lets teams spin up thousands of browser sessions on demand, with native SDKs in Python, Node, Rust, and Go.
Steel is designed for the realities of agent web use. Sessions start in under a second in region and can run up to twenty four hours, with cookies and local storage saved and injected so an agent can pick up where it left off. Built in rotating proxies, stealth configuration, a custom Chromium fork called Stealth Browser, dedicated IPs, and CAPTCHA solving keep automations from getting flagged, and agents can securely reach auth walled sites. A live Session Viewer with replays and Chrome DevTools debugging turns failed web interactions from guesswork into inspection, and Projects add per project isolation for sessions, credentials, and profiles. Steel Skills provides a catalog of reusable agent skills for the web, and scrape, screenshot, PDF, and structured extract endpoints turn pages into clean data.
Pricing keeps the open source build free to self host, offers free cloud credits to start, and adds Launch, Scale, and Enterprise plans with metered browser rates and usage credits, with an inactivity timeout that releases idle or crashed sessions so they stop billing. School researchers can email for free credit grants.
Vendor details
Canonical URL
https://steel.dev
Category
Agent infrastructure
Subcategory
Browser infrastructure
Funding status
Independent. Open source browser API with a public Docker container for self hosting, plus a managed cloud. Native SDKs in Python, Node, Rust, and Go, and an active web agent ecosystem.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
REST and WebSocket API with full Chrome DevTools Protocol, so Puppeteer, Playwright, and Selenium connect directly. Native SDKs in Python, Node, Rust, and Go, integration with agent frameworks like Hermes, and scrape, screenshot, PDF, and structured extract endpoints. Self host via Docker or run in Steel's cloud.
In practice
Your agent needs to browse real websites, not brittle scrapes. You create a Steel session over the API, connect Playwright via Chrome DevTools Protocol, and the agent navigates, clicks, and extracts with anti detection handled.
An agent keeps getting lost mid task and you cannot tell why. Steel's live Session Viewer with replays lets you watch the exact page state it saw, turning debugging into inspection instead of guessing.
Compliance requires keeping browsing data private. You self host the open source Steel container in your own environment, so sessions and cookies never leave your infrastructure.
Sources & related URLs
Capability coverage
8.0 / 14 capabilities · 57%
| Integrations & Tool CallingREST and WebSocket API, CDP, Puppeteer, Playwright, Selenium, SDKs in Python, Node, Rust, Go, docs 2026-07-06 | Full |
|---|---|
| Workflow OrchestrationSession lifecycle management and Steel Skills catalog of web agent skills, docs 2026-07-06 | Partial |
| Knowledge Grounding & RAGScrape, extract, and structured extract endpoints turn pages into data for RAG ingestion, docs 2026-07-06 | Partial |
| Human Oversight & GuardrailsNo guardrails or human oversight product | Unable to verify |
| Security, Identity & GovernanceIsolated browser sandbox, per project isolation for sessions, credentials, and profiles, self host for data control, docs 2026-07-06 | Full |
| Observability & AuditabilityLive Session Viewer with replays and Chrome DevTools debugging, docs 2026-07-06 | Partial |
| Memory & State PersistenceCookies and local storage saved and injected for persistent sessions, docs 2026-07-06 | Partial |
| Deployment & Data ResidencyManaged cloud plus open source self host via Docker for full data control, docs 2026-07-06 | Full |
| Prebuilt Agents, Templates & PacksSteel Skills catalog of reusable web agent skills and a cookbook, docs 2026-07-06 | Partial |
| Triggers & Channel CoverageNo first class event triggers or channel coverage | Unable to verify |
| Model Flexibility & RoutingModel agnostic browser infrastructure; does not provide or route models | Unable to verify |
| APIs, SDKs & MCP ExtensibilityREST and WebSocket API, CDP, open source, SDKs in Python, Node, Rust, Go, docs 2026-07-06 | Full |
| Testing, Debugging & OptimizationSession replays and live debugging support testing and iteration, docs 2026-07-06 | Partial |
| Browser & Computer UseCore product: real browser control for agents to navigate, click, fill, and screenshot the web, docs 2026-07-06 | Full |
Pricing
Open source self host free; free cloud credits; Launch, Scale, Enterprise (metered browser rates)
browser session usage and credits
Included quota
Open source self host is free (you pay only your own infrastructure). The managed cloud gives free credits to start, then Launch, Scale, and Enterprise plans meter browser session usage against plan credits and rates, with an inactivity timeout that releases idle sessions.
What is public
The open source self host path, free cloud credits, and plan tier names are public; exact metered plan prices are less transparent.
Billing mechanics
Free open source self host, or managed cloud plans (Launch, Scale, Enterprise) that meter browser session time and usage credits, with an inactivity timeout to stop billing idle or crashed sessions.
Cost watchouts
Metered browser session time drives cost, and long running or high concurrency agent workloads accrue quickly. Chromium is memory intensive when self hosting (roughly 200 to 500 MB per concurrent session), so self host infrastructure scales with concurrency.
Variable cost rationale
Cost is metered browser session time, so concurrency and session duration drive nearly all of it, though self hosting the open source build caps the vendor cost entirely.
Overage / add-ons
Cloud usage meters browser session time and credits against the plan; self host has no vendor metering beyond your own infrastructure.
Sales call required
Mixed (some tiers require a call)
Free / trial
Open source free to self host, plus free cloud credits to start
Lowest paid plan
Launch plan (metered browser rates), or free via self hosted open source
Commercial notes
Independent open source browser API for AI agents, with a self host Docker path and a managed cloud.
Key ambiguities
Exact plan prices for Launch and Scale are not captured here; confirm current metered rates on the pricing page.
Related vendors
- AgentOps — Agent observability and reliability platform with broad model and…
- Agno — High-performance agent runtime and framework (formerly Phidata) with…
- Apify — Cloud platform for web scraping and automation with 45,000+ prebuilt…
- Arcade — Authenticated tool calling platform and MCP runtime that handles…
- Arize AI — AI observability and evaluation platform that traces, evaluates, and…
- Beam — Open source serverless sandboxes and GPU cloud for production…