ConnectorsLive

MCP Server

Connect Claude — or any Model Context Protocol client — to SupaGamma and research prediction-market data right inside the conversation. Search markets, pull per-market trades, L2 orderbook and OHLCV, and the raw event tape — everything your subscription covers.

What you need

A paid Claude plan (Pro, Team, Max, or Enterprise) — custom connectors require one.
A SupaGamma account with an API key (its default download permission is what the connector uses).
A positive credit balance — top up on the billing page. Every tool except get_balance needs a balance, and paid pulls debit it.

One URL, every client

SupaGamma speaks OAuth 2.1 with dynamic client registration, so every client below connects with the server URL alone — no API key pasted into a config file, and no separate integration per vendor. You paste a download-scoped key once, on our own consent screen, and the client holds a short-lived token from then on.

https://mcp.supagamma.com/mcp

Set it up in your tool

Same URL everywhere — the file, the keys and the command differ. Jump to yours.

ClaudeWeb and desktop apps
Claude CodeTerminal, one command
Cursormcp.json
VS CodeGitHub Copilot
Codex CLIconfig.toml + oauth login
opencodeopencode.json
WindsurfCascade
ChatGPTConnectors, developer mode

Claude

Web and desktop apps

Click-through

Open Settings → Connectors.
Click Add custom connector and name it SupaGamma.
Paste the server URL below, then click Connect.
Approve on the SupaGamma consent screen with a download-scoped API key.
Enable the connector in a chat from the tools menu.

Server URL

https://mcp.supagamma.com/mcp

Claude Code

Terminal, one command

One command

Run the command below. Add -s user to make it available in every project rather than just the current one.
The first call opens your browser for consent — paste a download-scoped API key there.

claude mcp add --transport http supagamma https://mcp.supagamma.com/mcp

Cursor

mcp.json

Config file

Settings → MCP → Add new MCP server, or edit the file directly.
Use ~/.cursor/mcp.json for every project, or .cursor/mcp.json for just this one.

~/.cursor/mcp.json

{
  "mcpServers": {
    "supagamma": {
      "url": "https://mcp.supagamma.com/mcp"
    }
  }
}

VS Code

GitHub Copilot

Config file

Create .vscode/mcp.json in your workspace.
Click Start on the CodeLens that appears above the server entry.
Note the key is servers here, not mcpServers, and type is required.

.vscode/mcp.json

{
  "servers": {
    "supagamma": {
      "type": "http",
      "url": "https://mcp.supagamma.com/mcp"
    }
  }
}

Codex CLI

config.toml + oauth login

Config file

Add the block below to ~/.codex/config.toml. There is no codex mcp add for HTTP servers — only for stdio ones — so this is edited by hand.
Then run codex mcp login supagamma to complete the OAuth flow.

~/.codex/config.toml

[mcp_servers.supagamma]
url = "https://mcp.supagamma.com/mcp"
auth = "oauth"

opencode

opencode.json

Config file

Add a mcp entry to opencode.json.
The type must be "remote" — the default is a local stdio process.

opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "supagamma": {
      "type": "remote",
      "url": "https://mcp.supagamma.com/mcp",
      "enabled": true
    }
  }
}

Windsurf

Cascade

Config file

Edit ~/.codeium/windsurf/mcp_config.json, then refresh the server list in Cascade.
Windsurf expects serverUrl for remote servers, not url — this one trips people up.

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "supagamma": {
      "serverUrl": "https://mcp.supagamma.com/mcp"
    }
  }
}

ChatGPT

Connectors, developer mode

Click-through

Custom MCP connectors live under Settings → Connectors and require developer mode.
Availability depends on your plan, and OpenAI has moved this surface more than once — if you do not see it, check their current connector docs.
Add the server URL below and complete the OAuth consent.

Server URL

https://mcp.supagamma.com/mcp

Using something else? Any MCP client that supports remote servers over streamable HTTP with OAuth 2.1 will work with the same URL — that is the point of the protocol.

What it can do

Free tools return metadata and quotes and never touch your balance. Paid tools deliver data and charge per delivered row (the same per-MB rates as downloads), and are idempotent within a short window — a dropped response or a retry won't double-charge.

Tool	Cost	What it does
get_balance	free	Your remaining legacy credit balance (redeemable toward a plan).
search_markets	free	Find markets that have data for sale, ranked by trade count.
get_market	free	One market's detail — resolved outcome + data-availability window.
list_series	free	Venue / product data series with coverage windows.
estimate_cost	free	Quote a pull's row count + cost before you buy.
get_market_trades	paid	The trade tape for one market (both outcome tokens).
get_market_orderbook	paid	Reconstructed L2 orderbook snapshots for one market.
get_market_ohlcv	paid	OHLCV candles (1m / 1h / 1d) for one market.
get_l2_slice	paid	A slice of a venue-wide raw L2 / trade stream.

get_market_* tools take a market_id — use them when you name a market. get_l2_slice reads a venue-wide raw stream by id + time window. All deliveries are bounded; for full files use the REST download API.

Pricing

Call estimate_cost first to size a pull. Slices are bounded for inline research; bulk goes through the download API. See the pricing page for per-data-type rates.

Try it

“Search SupaGamma markets for bitcoin, ranked by trade count.”
“Get the trades for market 546814.”
“Estimate the cost of 1h OHLCV for that market, then pull a sample.”

Your key stays safe

Your API key is sent once over HTTPS, exchanged for a short-lived access token, and never stored by the connector. Revoke it any time in your dashboard and the connection stops working within ~15 minutes.