Connect Claude, Cursor, VS Code, or any agent to MeshTool's hosted MCP server. Your AI gains tools to analyze, personalize, extract, orchestrate, and run real work — over one endpoint. You don't build the model; you plug in the capabilities.
Grab an API key from app.meshtool.ai — a sk_tz_… token. It authorizes every call and meters usage.
Claude Desktop (and any stdio client) — via the mcp-remote bridge:
{
"mcpServers": {
"meshtool": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://api.meshtool.ai/mcp",
"--header", "Authorization: Bearer YOUR_KEY"]
}
}
}
Cursor / VS Code / clients with native remote MCP — point straight at the URL:
{
"mcpServers": {
"meshtool": {
"url": "https://api.meshtool.ai/mcp",
"headers": { "Authorization": "Bearer YOUR_KEY" }
}
}
}
Swap YOUR_KEY for your sk_tz_ key, restart the client — done.
Live catalog from the running server — the agent decides when to call each:
Every call needs a valid Bearer key (not just Origin checks — this is a public remote server). Keys are revocable and per-account, usage is metered, and each tool declares its exact input/output schema. Nothing runs without your key.
This is the B2B side of the cognitive-websites thesis: apps and agents that remember, think, and act — composed over MCP, "agents calling agents." Expose your own data/APIs as tools the same way.