Sato Hub

Model Context Protocol

Install the Sato Hub MCP server

Read-only, no key, no account. Your agent gets the scored index of what onchain agents are built from — frameworks, MCP servers, wallets, x402 rails, trading venues — rebuilt daily, each record with a page you can cite.

EndpointStreamable HTTP · read-only · no auth
https://satohub.ai/api/mcp

5 of these clients install in one click. The rest take a config block, and the block differs more than it should: the field for a remote server is url in some clients and serverUrl, httpUrl, uri or a nested type in others. Every block below is generated from the endpoint above and was read off that client’s own docs on 2026-09-13, so none of them is a guess at another client’s format.

Editors

Cursor

one click
Add to Cursor

Format read from their docs

VS Code

one click
Add to VS Code

The top-level key is `servers`, not `mcpServers`.

Format read from their docs

VS Code Insiders

one click
Add to VS Code Insiders

Format read from their docs

Windsurf

The key is `serverUrl`. Windsurf's own deep link resolves against its registry and cannot carry an arbitrary URL, so there is no one-click here yet.

Format read from their docs

Zed

The top-level key is `context_servers`. Settings → AI → MCP Servers → Add Remote Server does the same thing.

Format read from their docs

Cline

`streamableHttp` is camelCase here and hyphenated in Continue. They are not interchangeable.

Format read from their docs

Continue

`mcpServers` is a list here, and the type is hyphenated.

Format read from their docs

JetBrains AI Assistant

Pasted into the dialog — there is no user-editable config file.

Format read from their docs

CLI

Claude Code

claude mcp add --transport http satohub https://satohub.ai/api/mcp

`type: "http"` is required — a bare `url` is read as a stdio config.

Format read from their docs

Codex CLI

codex mcp add satohub --url https://satohub.ai/api/mcp

Format read from their docs

Gemini CLI

gemini mcp add --transport http --scope user satohub https://satohub.ai/api/mcp

The key is `httpUrl`. A plain `url` here means SSE, which this endpoint does not serve.

Format read from their docs

Desktop & terminal

LM Studio

one click
Add to LM Studio

The deep link needs LM Studio 0.3.17 or newer.

Format read from their docs

Goose

one click
Add to Goose

The deep link passes `url`; the config file wants `uri`. Goose has an open report of streamable-HTTP deep links failing quietly — if nothing appears, use `goose configure` and the block above.

Format read from their docs

Claude Desktop

Paste the URL, not a JSON block — `claude_desktop_config.json` takes local servers only. The connection is made from Anthropic's servers, not your machine. On the free plan you get one custom connector.

Format read from their docs

Warp

Warp's docs describe a `url` server as SSE. This endpoint serves Streamable HTTP and has SSE disabled, so we have not confirmed Warp connects. Reported as untested rather than left to fail quietly.

Format unconfirmed — see their docs

Then tell the agent to use it

Installing the server does not make a model reach for it. A coding agent asked which wallet SDK works on a given chain will answer from recall unless something in the repo says otherwise. The agent kit is that something — a drop-in section for the rules file your client already reads.

No client at all

The same data is free over plain HTTP: /api/export/index.json (also .csv and .ndjson), /api/changes for what moved, and /api/openapi.json for every public operation. This page itself is machine-readable at /api/install.json.

There is no Claude Desktop extension bundle. The bundle format spawns a local process and has no remote server type, so a “remote” bundle could only wrap a third-party bridge package fetched at install time. Claude Desktop takes the URL directly under Connectors, which is the better path, so we did not ship the worse one.

What the tools return and what each one is for: /mcp.