OpenTabs

Why FADING

FADING (64) — genuinely impressive engineering and the most useful automation approach available (real internal APIs beat DOM or screenshot driving), with a clever self-improving plugin generator. Held down by risk the surface does not price in: undocumented APIs that can break without notice, probable terms-of-service exposure borne by the user's own account, financial and infrastructure targets among the plugins, and no visible permission model or destructive-action confirmation for an extension acting as you across every logged-in tab.

Public-surface checklist

• PASS Homepage loads (required)
• PASS Primary value prop (required) — 'Every web app is an API'
• PASS Cta present (required) — 'Get Started' / npm install -g @opentabs-dev/cli
• FAIL Pricing or access — No pricing on the captured surface
• PASS Evidence or demo — Mechanism explained step by step and plugin catalogue shown; no third-party review

What we saw

4 screenshots captured by the Hlido engine during the reviewed run (run-a61bc9480b979b61-opentabs-dev). Our own captures — not vendor marketing material.

×

←→

What it does well

• Calls real internal APIs rather than driving the DOM or screenshots — far more reliable automation
• Very large claimed surface: 100+ plugins, ~2,000 tools across mainstream services
• Uses the existing authenticated session, so no separate credential handling or API keys
• Self-improving: point the AI at a site and it discovers the APIs and scaffolds the plugin
• Works with any MCP client supporting Streamable HTTP (Claude Code, Cursor, Windsurf, OpenCode)
• Explicit about the mechanism — it does not pretend to be an official integration

What it fails at

• Depends on undocumented internal APIs that can change or break with no notice or deprecation path
• Likely terms-of-service exposure for the sites automated — and the risk sits with the user's account, not the vendor
• No permission model, per-plugin scoping, or destructive-action confirmation visible on the captured surface
• An extension able to act as you across every logged-in tab is an extremely broad trust grant
• Financial and infrastructure targets (Stripe, AWS, Robinhood) raise the cost of an unintended call
• Plugin quality across 100+ community plugins will be highly uneven; most were AI-generated in minutes
• No adoption figures or third-party security review

Pricing & access

• Pricing findable on the public surfaceFAIL No pricing on the captured surface (tested 2026-08-07)

Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-08-07.

Related agents

• Anyscale Infrastructure · 90
• Fireworks AI Infrastructure · 90
• Hyperbolic Infrastructure · 90

Agent relevance

API CLI MCP SDK Behavioral-testable

Agentic-Commerce Readiness 72/100 · INTEGRABLE

Independent readiness for agent delegation & transaction. How it’s scored · check live

Built entirely for agents: an MCP server over Streamable HTTP exposing ~2,000 tools, with a CLI to start it. The agent issues normal MCP tool calls and the extension executes them in a real browser tab.

Agent-friendly score: 9/10

Evidence

• Chrome extension plus MCP server calling web apps' internal APIs via the user's session — source (2026-08-07) verified
• 100+ plugins, ~2,000 tools across Slack, GitHub, Jira, Notion, Figma, AWS, Stripe and 90+ more — source (2026-08-07)
• Explicitly 'no screenshots, no DOM, no guessing' — real backend calls — source (2026-08-07) verified
• AI-generated plugins: point it at a site and it discovers APIs and scaffolds the plugin — source (2026-08-07) verified
• A permission model, per-plugin scoping, or destructive-action confirmation — source (2026-08-07)