Web Speed offers MCP-native sitemap registry to speed up AI agents
Web Speed is an open-source, MCP-native tool that parses web pages into lightweight sitemaps for AI agents, reducing the cost and latency of browser-control tasks by eliminating full HTML or screenshot analysis.
Score breakdown
Teams building MCP-based browser agents can reduce token consumption and latency by swapping full-page HTML parsing for Web Speed's pre-parsed sitemap format, with further gains available through the shared cache for commonly visited sites.
- 01Web Speed parses web pages into compact, agent-readable sitemaps instead of passing raw HTML or screenshots to AI agents.
- 02The tool is MCP-native, making it compatible with any AI that supports the MCP protocol for browser control.
- 03A global shared cache stores sitemaps uploaded by users so future agents can retrieve pre-parsed versions of the same pages.
Web Speed is an MCP-native, open-source tool built to make AI browser agents faster and cheaper. Instead of feeding an agent the full raw HTML of a page or a screenshot, Web Speed parses the page into a concise, structured sitemap that the agent can read directly. By stripping away the noise of full HTML, the tool reduces the amount of tokens an agent must process, lowering both cost and latency for web-navigation tasks.
The project ships in two open-source variants on GitHub — a standard version and a post-authentication version for pages that require a login — alongside a hosted API with documentation at getwebspeed.io.
The project ships in two open-source variants on GitHub — a standard version and a post-authentication version for pages that require a login — alongside a hosted API with documentation at getwebspeed.io. The most distinctive aspect of the project is its global sitemap cache: whenever a user visits a site through Web Speed, the resulting sitemap is pushed to a shared server. Subsequent agents querying the same site can pull the cached sitemap rather than re-parsing the page from scratch. Access to this shared cache is currently gated behind the paid API tier.
Key facts
- 01Web Speed parses web pages into compact, agent-readable sitemaps instead of passing raw HTML or screenshots to AI agents.
- 02The tool is MCP-native, making it compatible with any AI that supports the MCP protocol for browser control.
- 03A global shared cache stores sitemaps uploaded by users so future agents can retrieve pre-parsed versions of the same pages.
- 04Access to the shared cache is currently limited to the paid API version.
- 05Two open-source GitHub repositories are available: a standard version and a post-authentication version for login-gated pages.
- 06The project is posted by Dominic_P and is seeking community feedback, ideas, and criticism.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 8, 2026 · 15:36 UTC. How this works →