Guide

Hosted MCP server for Claude, Cursor, and Windsurf

Instadash runs a hosted, OAuth-authenticated MCP server at mcp.instadash.io. Paste it into Claude Desktop, Cursor, or any MCP client; your agent gains tools to push JSON, read rows, query, and search structured data without you running infrastructure.

What MCP is, in one paragraph

The Model Context Protocol is a standard way for an AI agent to discover and call external tools. The host (Claude Desktop, Cursor, Windsurf) connects to MCP servers; each server advertises a list of tools the model can invoke. Tools are typed, schema-validated, and accept structured input — much cleaner than ad-hoc HTTP fetching.

The Instadash MCP server

Endpoint:

https://mcp.instadash.io/mcp

Auth: OAuth 2.0 (the client handles the browser redirect for you).

Transport: Streamable HTTP — no local binary to install, no Node version pinning. Works the same on Mac, Linux, Windows, and remote IDE sessions.

Setup: Claude Desktop

Open the Claude Desktop config (Settings → Developer → Edit Config) and add:

{
  "mcpServers": {
    "instadash": {
      "type": "http",
      "url": "https://mcp.instadash.io/mcp"
    }
  }
}

Restart Claude Desktop. The first call to an Instadash tool will pop a browser window for OAuth. Approve, and tool calls land instantly thereafter.

Setup: Cursor

Cursor's MCP config lives in ~/.cursor/mcp.json (or the workspace .cursor/mcp.json):

{
  "mcpServers": {
    "instadash": {
      "url": "https://mcp.instadash.io/mcp"
    }
  }
}

Reload Cursor; the Composer agent will pick up the tools.

Setup: Windsurf

Settings → Cascade → MCP Servers → Add. Paste:

https://mcp.instadash.io/mcp

Save. OAuth on first call, same as the others.

Setup: Claude Code (CLI)

claude mcp add instadash https://mcp.instadash.io/mcp

The CLI persists the config to the current workspace.

What you can now do

Once connected, the agent can:

  • instadash_push — create or update a grid with structured rows.
  • instadash_read — fetch rows with pagination, filter, and version selection.
  • instadash_get_schema — inspect column types before pushing.
  • instadash_list — list every grid in the workspace.
  • instadash_edits — read human-touched cells (HITL read-back).
  • instamesh_search — semantic + keyword search across all public grids.

Example agent prompts:

"Scrape these 20 SaaS pricing pages and push the results to a grid called pricing-research."

"Read the latest version of grid acme/customer-feedback, group by sentiment, and tell me the top three complaint themes."

"Search the public mesh for grids about MCP servers."

Each prompt expands into one or two MCP tool calls. No code on your side.

Where to go next

  • Full MCP doc: /docs/mcp
  • The mesh protocol (semantic search across grids): /docs/mesh-protocol
  • Auth / OAuth deep dive: /docs/auth

FAQ

What MCP transport does Instadash use?

Streamable HTTP over HTTPS at https://mcp.instadash.io/mcp. This is the modern transport — no local stdio binary, no Node-version mismatches, and it works in remote IDEs and web-based clients.

Is the MCP server free?

Yes. The MCP server itself is free to use. You only pay if your grid usage exceeds the free tier on rows or stored bytes — see /pricing.

Does the server work with Anthropic's Claude Code CLI?

Yes — Claude Code reads the same MCP config format as Claude Desktop. Add the server URL once and every Claude Code session in that workspace can call the tools.

How is auth handled?

OAuth 2.0. The first time the MCP client connects, you're redirected to instadash.io to sign in and approve. After that, the client refreshes tokens automatically. No API key paste.

Which MCP tools are exposed?

instadash_push, instadash_read, instadash_get_schema, instadash_list, instadash_list_versions, instadash_edits, instamesh_search, and a few admin tools. The full list is at https://instadash.io/.well-known/mcp.json.

Related

/docs/mcp →/docs/mesh-protocol →/docs/auth →/recipes/claude-mcp-instadash →

More guides

Mesh-augmented generation vs RAG, explained
Retrieval-augmented generation pulls unstructured text into an LLM prompt. Mesh-augmented generation does the same for structured data — agents query a public mesh of live, queryable grids that are schema-typed, citation-bearing, and addressable by URL, instead of scraping or re-prompting.
Turn JSON, JSONL, or CSV into an API for AI agents
You have JSON, JSONL, or CSV and you want it accessible to AI agents as an API. Instadash turns any structured file into a hosted grid with a JSON API, search index, and MCP endpoint — one CLI push command, no server to deploy.

Related across sections

Use case
Persistent storage for AI agent structured output
AI agents that extract or generate structured data need somewhere to put it. Instadash gives the agent a hosted grid at a stable URL — push JSON or JSONL once, then read it back via API, MCP, or markdown without re-running the agent or re-streaming tokens.
Use case
Human-in-the-loop review for AI agent output
Some agent outputs need a human to approve before action — outbound emails, refunds, ticket triage. Instadash gives you an editable shared grid: the agent writes rows, humans tick approval columns, the agent reads back exactly which rows changed and acts.
Comparison
Instadash vs Airtable, Fastio, and Datasette
Choosing where AI agents store structured data. Airtable optimizes for human teams; Datasette publishes SQLite as a queryable site; Fastio focuses on developer-friendly APIs; Instadash is built MCP-first for AI agents — agents push from any framework and read via MCP, JSON, or markdown.
all docshomeedit on github

canonical: /guides/hosted-mcp-data-grid