WebMCP: Agent Tools on AI Spectrum
Documentation for AI agents on the WebMCP tools this site exposes: searchBlog, compareModels, and getModelInfo, with schemas and example calls.
This site exposes browser-side tools via the WebMCP API (navigator.modelContext) for agents running in a real browser context (e.g. Playwright, Comet, Atlas). Agents that can only fetch HTML without executing JS can read the static manifest instead.
Available tools
searchBlog
Full-text search over AI Spectrum posts and pages.
| Param | Type | Description |
|---|---|---|
query | string | Keyword(s) to search across post titles, descriptions, and body text |
compareModels
Compare two AI models by their HLE (Humanity's Last Exam) and GPQA Diamond benchmark scores.
| Param | Type | Description |
|---|---|---|
model1 | string | Name of the first model, e.g. 'Claude Opus 5' |
model2 | string | Name of the second model, e.g. 'GPT-5.4' |
getModelInfo
Get a single AI model's HLE and GPQA Diamond benchmark scores.
| Param | Type | Description |
|---|---|---|
model | string | Name of the model, e.g. 'Claude Opus 5' |
All three tools are read-only (readOnlyHint: true) and require no authentication.
Calling the tools from a browser-executing agent
const tools = await navigator.modelContext.listTools();
const result = await navigator.modelContext.executeTool("searchBlog", { query: "WebMCP" });
Static manifest for non-JS agents
If your agent only fetches HTML, read the schemas directly instead of executing JS:
/webmcp-manifest.json/.well-known/webmcp.json(same content, conventional path)- Inline JSON-LD (
#webmcp-manifest) embedded on every page /llms.txtfor a plain-text summary
