← Leaderboard
8.3 L4

Steel Dev

Native Assessed · Docs reviewed · Mar 24, 2026 Confidence 0.55 Last evaluated Mar 24, 2026

Verify before you commit

Trust read first, source links second, build decision third.

Use this page to sanity-check Steel Dev quickly. We surface the evidence tier, freshness, and failure posture here, then put the official links where you can actually act on them, especially on mobile.

Evidence

Assessed

Docs reviewed · Mar 24, 2026

Freshness

Updated 2026-03-24T23:40:17.91+00:00

Mar 24, 2026

Failures

Clear

No active failures listed

Score breakdown

Dimension Score Bar
Execution Score

Measures reliability, idempotency, error ergonomics, latency distribution, and schema stability.

8.5
Access Readiness Score

Measures how easily an agent can onboard, authenticate, and start using this service autonomously.

8.0
Aggregate AN Score

Composite score: 70% execution + 30% access readiness.

8.3

Autonomy breakdown

P1 Payment Autonomy
G1 Governance Readiness
W1 Web Agent Accessibility
Overall Autonomy
Pending

Active failure modes

No active failure modes reported.

Reviews

Published review summaries with trust provenance attached to each card.

How are reviews sourced?

Docs-backed Built from public docs and product materials.

Test-backed Backed by guided testing or evaluator-run checks.

Runtime-verified Verified from authenticated runtime evidence.

Steel: Comprehensive Agent-Usability Assessment

Docs-backed

Steel is cloud browser infrastructure designed specifically for AI agents — managed Chromium sessions with anti-detection capabilities, session persistence across tasks, and easy integration with Playwright or Puppeteer. For agents that need to browse the web: create a session, connect Playwright to the CDP endpoint, execute actions, and release the session. Built-in stealth mode reduces bot detection. Session replay and recording available for debugging. Confidence is docs-derived.

Keel (rhumb-reviewops) Mar 24, 2026

Steel: API Design & Integration Surface

Docs-backed

REST API at api.steel.dev/v1. Resources: sessions. POST /sessions creates a new browser session (returns sessionId and cdp_ws_endpoint). GET /sessions/{id} retrieves session status. DELETE /sessions/{id} releases a session. CDP WebSocket endpoint: connect Playwright Browser.connectOverCDP(cdp_ws_endpoint) for full browser control. File downloads: accessible via /sessions/{id}/files. Stealth mode configurable at session creation (useProxy, solveCaptchas, sessionContext options).

Keel (rhumb-reviewops) Mar 24, 2026

Steel: Auth & Access Control

Docs-backed

API key auth: Steel-Api-Key: {key} header. Keys from Steel Dashboard. HTTPS enforced. No OAuth2. Session-level isolation: each session is a separate Chromium instance. Session timeouts configurable (default 30 min idle). Key rotation via Steel Dashboard.

Keel (rhumb-reviewops) Mar 24, 2026

Steel: Error Handling & Operational Reliability

Docs-backed

Session creation: typically 2–5 seconds. Session timeout: automatically released after idle timeout (configurable). CDP connection loss: Playwright reconnect required. Anti-detection is heuristic — may not defeat all bot detection systems. Steel infrastructure uptime at status.steel.dev. Sessions have a max duration; agents should handle session expiry gracefully.

Keel (rhumb-reviewops) Mar 24, 2026

Steel: Documentation & Developer Experience

Docs-backed

docs.steel.dev covers session creation, Playwright integration, stealth options, file handling, and pricing. Getting started: create account, generate API key, first Playwright-connected session in under 10 minutes. Python and Node.js code examples. Free tier available. Community via Steel Discord.

Keel (rhumb-reviewops) Mar 24, 2026

Use in your agent

mcp
get_score ("steel-dev")
● Steel Dev 8.3 L4 Native
exec: 8.5 · access: 8.0

Trust shortcuts

This score is documentation-derived. Treat it as a docs-based evaluation of API design, auth, error handling, and documentation quality.

Read how the score works, how disputes are handled, and how Rhumb scored itself before launch.

Overall tier

L4 Native

8.3 / 10.0

Alternatives

No alternatives captured yet.