Method

Public docs, dated, and narrow on purpose.

BrowseBoard compares tools whose job is to let a model drive a browser. The grid is the coordinator seed checked on 2026-10-19. Every row keeps that date and its source URLs.

How a cell gets filled

The file is data/browseboard-seed.json.

All fields gathered from public docs/repos only. Prefer unknown over guessing. Prices and capabilities may change; re-check sources before publishing.

A missing or conflicting public page stays unknown. Prices are the figures on the linked pages, not estimates. Skyvern's billing table is the plan source when the FAQ wording drifts.

MCP marked partial is not a yes. Stagehand's in-repo adapter is experimental, and OpenAI Computer Use is a Responses API tool set rather than a Playwright-style MCP server.

What stays off the board

Cloud browsers sold by the hour are infrastructure, not peers.

Browserbase (infra companion — not a separate agent row). Do not render as a peer agent row; Stagehand is the agent/framework row. Cloud browser pricing — out of BrowseBoard primary scope (see scrapingapiprices / Browser Arena territory). Marked unknown here intentionally.

LaVague. Third-party comparisons note active development ended ~Jan 2025; not seeded as an active row without confirming current maintenance.
AgentQL. Semantic query/extraction layer over Playwright, not a full autonomous browser-agent framework — out of primary matrix scope.
Pure cloud browser hosts (Steel, Browserless, Hyperbrowser, etc.). Infra $/hour layer — deliberately out of BrowseBoard scope.

Disclaimer

This is a reading aid, not legal, security, or purchasing advice. AGPL, SSPL, credential vaults, and anti-bot claims need your own review. Vendor stealth scores are vendor benches, cited as claims.

BrowseBoard does not run these agents and does not rank them. Re-fetch the pricing pages before you treat a number as current.