Tresca for AI agents
Tresca's MCP server lets your AI agent look up the 434 MCP servers and agent skills on Tresca for itself: which exist for a tool, which are official, what each can do and how to install it. It is read-only and needs no sign-in.
Connect
Add https://tresca.app/mcp to your AI client as a remote MCP server (streamable HTTP). For example:
Claude Code
Run in a terminal (Claude Code's documentation)
claude mcp add --transport http tresca https://tresca.app/mcpCursor
.cursor/mcp.json or ~/.cursor/mcp.json (Cursor's documentation)
{
"mcpServers": {
"tresca": {
"url": "https://tresca.app/mcp"
}
}
}Visual Studio Code
.vscode/mcp.json (Visual Studio Code's documentation)
{
"servers": {
"tresca": {
"type": "http",
"url": "https://tresca.app/mcp"
}
}
}Codex
~/.codex/config.toml (Codex's documentation)
[mcp_servers.tresca]
url = "https://tresca.app/mcp"In Claude Desktop, add it under Customize > Connectors as a custom connector with the URL above. Other clients have a similar setting for a remote server.
What it can answer
get_tool: Is there an MCP server or skill for a tool, is one official, and does the vendor have an AI assistant?search_listings: Search by words, or browse by tool, discipline or AI client; filter by type, official, operating system and whether it can change modelsget_listing: Every fact for one listing, with its sources and the date it was checkedget_install_steps: Install steps copied exactly from the project's documentation, each with its sourceget_vendor_assistant: A vendor's AI assistant: status in each product in the vendor's words, and how it relates to MCPlist_disciplines: Disciplines and the tools in each, with counts
Answers come from the same data as the site's pages and change when the site does. Results follow the site's order, official first, and nothing is ranked. Tresca records what each project's documentation says; it doesn't audit code, so your agent should show you the facts and ask before installing anything. AI-generated models, drawings and results must be checked by a qualified engineer before use.
Limits and privacy
Up to 60 requests a minute from one address. The server doesn't store the questions agents ask; see the privacy page.
Prefer files? The machine-readable index and llms.txt list every listing. Something wrong or missing? Suggest a listing or an edit.