3102 Model Context Protocol servers — searchable, rated, ready to install.
Automatic project memory for Claude Code and Cursor; Codex and Claude Desktop over MCP.
How ChatGPT, Claude, Gemini and Perplexity read a site: free check, fix plan, compare, history.
AI brand monitor: what ChatGPT and Claude say about your brand, press tracking, executive briefs.
Convert JSON samples into TypeScript interfaces and Zod schemas, with inference caveats.
Prove you made it first. Blockchain Timestamp creative work via Claude or any MCP-enabled AI Agent.
Your own closet in your AI: outfits from clothes you own, rendered on a model inline.
Model, measure and verify parametric parts and assemblies in a running SolidWorks (Windows).
Connect Claude to WordPress safely: signed requests, scoped keys, approvals, logs, snapshots.
WordPress MCP server plugin for Claude, ChatGPT and more. Reads by default, one switch to write.
Scans LLM/agent code for Unbounded Consumption / denial-of-wallet risks (OWASP LLM Top 10 2026 #6)
MCP server exposing Actual Budget accounts, transactions, budgets and reporting to LLM assistants.
AI fashion design MCP: generate concepts, models, fabrics & looks live on your StyTrix canvas.
See whether ChatGPT, Claude and Gemini recommend a startup's product, and who they name instead.
Live model catalog with prices and measured uptime, rankings, credits and spend, spend-capped chat.
Reads .mcp.json / .vscode/mcp.json / claude_desktop_config.json while you edit it and marks the line
Run DeepInfra inference, list models and read account rate limits.
Read your recorded AI agent runs, and what changed when you swapped a model or prompt.
Search, organize, and chat with your saved Reddit posts from Claude, Cursor, and any MCP client.
Generate and edit images with every image model on the platform. Returns the image to the agent.
Pay-per-call LLM chat plus data tools: DNS, crypto, Base gas, Wikipedia. x402 USDC or Lightning.
Official DeepMind blog: Gemini & AI model releases. Register in-session — free testnet funds.
Synthetic checks, nightly regression replay and model-drift alerts for AI agents
MCP server caching expensive LLM responses: HTTP-style freshness plus semantic recall.
Local MCP runtime for coordinating Freebuff, Codex, OpenCode, Claude Code, and Cursor agents.
POST /mcp