model-context

Talk with your notes in Claude. RAG over your Apple Notes using Model Context Protocol.

Coral is an open-source, local-first session bridge that lets AI agents borrow a user's already-authenticated browser sessions on a per-site, per-action, fully audited basis.

🚀 AI-powered freelance gig aggregator MCP server for Claude Desktop. Search gigs, generate proposals, negotiate rates & optimize profiles with ChatGroq AI.

Quick-reference guide to understanding common tech protocols — what they do and how to choose the right ones for tasks like secure logins, file transfer, video output, and building AI workflows. Includes comparisons of web, device, and model interaction protocols.

Expose a page's WebMCP tools as a real MCP server: an in-page JSON-RPC handler (initialize/tools/list/tools/call) over a pluggable transport.

Zero-config highlighting of the DOM changes an AI agent's WebMCP tool calls make — overlay rings, change timeline, off-screen reveal. Framework-agnostic core + React binding.

CLI based MCP client.

React hooks that register/unregister WebMCP tools with component lifecycle — the agent only sees tools for what's currently on screen.

Client-side human-in-the-loop for WebMCP: gate agent tool calls behind an approval policy (auto/confirm/deny), with a built-in modal or your own approver.

Define, validate, and register structured browser tools for the experimental WebMCP API

Qualify WebMCP interoperability against frozen profiles: qualified / not_qualified / insufficient_evidence - no execution.

Type-safe defineTool for WebMCP: infer execute() arg types from a JSON Schema, with optional runtime validation. Zero deps.

Measure WebMCP providers across models and runtimes. Attribute every failure to the layer that caused it. Keep the provider blameless when it's not at fault.

Model-agnostic runtime for executing stateful, multi-step trajectories across independent WebMCP providers.

Zero-config dev panel for WebMCP: lists a page's tools, auto-builds a form from each inputSchema, invokes them, shows results. Pairs with webmcp-highlight.

Test WebMCP tools without a flagged Chrome or a live model: a mock navigator.modelContext plus a driver to list, validate-args, invoke, and snapshot the toolset.

A neutral execution-control seam for agent tool calls: allow / block / indeterminate, only allow proceeds.

Client-side drift detection for a WebMCP toolset: hash the tools, freeze a baseline, and get notified when what a page exposes to agents changes.

A navigator.modelContext polyfill for WebMCP: installs a spec-faithful shim only when the browser lacks it, so you can develop and demo on any browser.