OpenFinance Trading MCP Server

The OpenFinance MCP server supercharges AI agents with trading skills. Connect Claude, ChatGPT, Cursor, Codex or any MCP client and your agent can:

• Trade crypto spot and perpetuals on Hyperliquid, with limit, market and TWAP orders, leverage and margin control.

• Trade stocks, indices, commodities, forex and pre-IPO names as Hyperliquid HIP-3 perps (for example xyz:NVDA).

• Trade prediction markets on Polymarket: sports, politics, crypto and world events, plus trader leaderboards.

• Swap and bridge anything onchain across 40+ EVM chains and Solana, with Relay and Jupiter routing.

• Find and trade memecoins with token search across 200+ chains, live prices and liquidity.

• Launch tokens on Robinhood Chain and Solana, including tokens paired against tokenized stocks.

• Fund the account with fiat, through card, bank transfer or UPI.

There is nothing to install or host. The server runs at one HTTPS endpoint, and your agent connects to it like any other remote MCP server.

https://api.openfinance.tech/agent/mcp

Contents

• Quick start

• What your agent can do

• Supported chains

• Authentication

• Install in your AI client

• Use it from your own code

• Tool reference

• Example prompts

• Safety and security

• Pair it with OpenFinance Skills

• Verify the connection

• Troubleshooting

• FAQ

• Links

Related MCP server: @y0exchange/mcp

Quick start

Easy trading for your agent in under a minute.

1. Create your account. Sign in at openfinance.tech. Your trading wallets (one EVM address for every EVM chain, plus a Solana address) are created for you.

2. Add the server to your AI client. In Claude Code, for example:

claude mcp add --transport http openfinance-tech https://api.openfinance.tech/agent/mcp

3. Sign in and approve. The first time your agent uses the server, your client opens an OpenFinance sign-in page. Approve the connection on the consent screen and you're done.

Then ask your agent something like:

"What's BTC's funding rate on Hyperliquid right now, and how has it moved over the last day?"

Using a different client? See Install in your AI client.

What your agent can do

Supported chains

Authentication

Every call to the server is authenticated. You can sign in two ways; both act on your own OpenFinance account.

OAuth sign-in. The server follows the MCP authorization spec, so supporting clients discover it automatically. When you connect, you sign in to OpenFinance and see which account is being connected; you have to confirm twice before anything is authorized.

API key. Sign in at openfinance.tech, open External AI in the sidebar, and copy the API key shown under Connect OpenFinance to your AI. Keys start with open_.

The examples below read the key from an environment variable:

export OPENFINANCE_API_KEY="open_..."   # add to your shell profile, not to a repo

Install in your AI client

Use the server name openfinance-tech and the URL https://api.openfinance.tech/agent/mcp everywhere. Each section shows the OAuth setup first, then the API-key alternative.

OAuth (recommended)

claude mcp add --transport http openfinance-tech https://api.openfinance.tech/agent/mcp

Start Claude Code, run /mcp, choose openfinance-tech and select Authenticate. Add --scope user to make it available in every project.

API key

claude mcp add --transport http openfinance-tech https://api.openfinance.tech/agent/mcp \
  --header "x-api-key: $OPENFINANCE_API_KEY"

Shared project config (.mcp.json): Claude Code expands environment variables, so the key never lands in the file:

{
  "mcpServers": {
    "openfinance-tech": {
      "type": "http",
      "url": "https://api.openfinance.tech/agent/mcp",
      "headers": { "x-api-key": "${OPENFINANCE_API_KEY}" }
    }
  }
}

1. Open Settings → Connectors and click Add custom connector.

2. Name: OpenFinance. URL: https://api.openfinance.tech/agent/mcp.

3. Click Add, then Connect, and approve the connection on OpenFinance.

4. In a chat, open the tools menu and switch OpenFinance on.

Connectors you add on Claude.ai also appear in Claude Desktop. Custom connectors are available on Claude's paid plans; on Team and Enterprise an owner may need to add them first.

API key alternative (Claude Desktop only). Bridge the server through mcp-remote in claude_desktop_config.json (Settings → Developer → Edit Config), then fully quit and reopen Claude:

{
  "mcpServers": {
    "openfinance-tech": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://api.openfinance.tech/agent/mcp",
        "--header", "x-api-key:${OPENFINANCE_API_KEY}"
      ],
      "env": { "OPENFINANCE_API_KEY": "open_..." }
    }
  }
}

1. Open Settings → Apps & Connectors. If you don't see a Create button, turn on Developer mode under Advanced settings.

2. Click Create. Name: OpenFinance. MCP server URL: https://api.openfinance.tech/agent/mcp. Authentication: OAuth.

3. Save, sign in to OpenFinance and approve the connection.

4. In a new chat, pick OpenFinance from the tools or connectors menu.

Custom connectors need a ChatGPT plan that supports them (Plus, Pro, Business, Enterprise or Edu); workspace admins may need to allow them.

Click Add to Cursor, or add this to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project):

{
  "mcpServers": {
    "openfinance-tech": {
      "url": "https://api.openfinance.tech/agent/mcp"
    }
  }
}

Cursor shows Needs login next to the server. Click it to sign in to OpenFinance.

API key

{
  "mcpServers": {
    "openfinance-tech": {
      "url": "https://api.openfinance.tech/agent/mcp",
      "headers": { "x-api-key": "${env:OPENFINANCE_API_KEY}" }
    }
  }
}

Click Install in VS Code, or run MCP: Add Server from the Command Palette and paste the URL. VS Code signs you in with OAuth.

API key, with VS Code prompting for the key once and storing it securely (.vscode/mcp.json or your user mcp.json):

{
  "inputs": [
    {
      "type": "promptString",
      "id": "openfinance-api-key",
      "description": "OpenFinance API key",
      "password": true
    }
  ],
  "servers": {
    "openfinance-tech": {
      "type": "http",
      "url": "https://api.openfinance.tech/agent/mcp",
      "headers": { "x-api-key": "${input:openfinance-api-key}" }
    }
  }
}

VS Code uses servers, not mcpServers. Open Copilot Chat, switch to Agent mode, and the OpenFinance tools appear in the tools picker.

codex mcp add openfinance-tech --url https://api.openfinance.tech/agent/mcp

If Codex asks you to sign in, run codex mcp login openfinance-tech.

API key in ~/.codex/config.toml, read from your environment:

[mcp_servers.openfinance-tech]
url = "https://api.openfinance.tech/agent/mcp"
env_http_headers = { "x-api-key" = "OPENFINANCE_API_KEY" }

The CLI and the IDE extension share this config. Run /mcp in Codex to check the server is connected.

Add to ~/.gemini/settings.json:

{
  "mcpServers": {
    "openfinance-tech": {
      "httpUrl": "https://api.openfinance.tech/agent/mcp"
    }
  }
}

Then run /mcp auth openfinance-tech to sign in. For an API key instead, add "headers": { "x-api-key": "$OPENFINANCE_API_KEY" }.

Open Cascade, click the hammer icon, then Configure, and add:

{
  "mcpServers": {
    "openfinance-tech": {
      "serverUrl": "https://api.openfinance.tech/agent/mcp",
      "headers": { "x-api-key": "${env:OPENFINANCE_API_KEY}" }
    }
  }
}

Windsurf uses serverUrl rather than url. Refresh the server list after saving.

In the Cline panel, open MCP Servers → Configure MCP Servers and add:

{
  "mcpServers": {
    "openfinance-tech": {
      "type": "streamableHttp",
      "url": "https://api.openfinance.tech/agent/mcp",
      "headers": { "x-api-key": "open_..." }
    }
  }
}

Cline uses streamableHttp as the transport type. This file is stored on your machine; don't commit it.

Settings → Connectors → + Add Connectors → Custom MCP → + Add Custom MCP → Direct Configuration:

• Name: OpenFinance

• Transport: Streamable HTTP

• Server URL: https://api.openfinance.tech/agent/mcp

• Header: x-api-key: open_...

hermes mcp add openfinance-tech --url https://api.openfinance.tech/agent/mcp

Or in ~/.hermes/config.yaml:

mcp_servers:
  openfinance-tech:
    url: "https://api.openfinance.tech/agent/mcp"
    headers:
      x-api-key: "open_..."

Run /reload-mcp, or restart the agent.

In ~/.openclaw/openclaw.json (note the nested mcp.servers key):

{
  "mcp": {
    "servers": {
      "openfinance-tech": {
        "url": "https://api.openfinance.tech/agent/mcp",
        "headers": { "x-api-key": "open_..." }
      }
    }
  }
}

If your client only supports local (stdio) servers, bridge it with npx -y mcp-remote https://api.openfinance.tech/agent/mcp.

Use it from your own code

Building your own agent? Connect with the official MCP SDKs.

npm install @modelcontextprotocol/sdk

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";

const transport = new StreamableHTTPClientTransport(
  new URL("https://api.openfinance.tech/agent/mcp"),
  { requestInit: { headers: { "x-api-key": process.env.OPENFINANCE_API_KEY! } } },
);

const client = new Client({ name: "my-trading-agent", version: "1.0.0" });
await client.connect(transport);

const { tools } = await client.listTools();
console.log(tools.map((t) => t.name));

// Read-only: mid prices for every Hyperliquid market
const mids = await client.callTool({
  name: "hyperliquid",
  arguments: { action: "get_all_mids" },
});
console.log(mids.content);

pip install mcp

import asyncio, os
from mcp import ClientSession
from mcp.client.streamable_http import streamablehttp_client

URL = "https://api.openfinance.tech/agent/mcp"

async def main():
    headers = {"x-api-key": os.environ["OPENFINANCE_API_KEY"]}
    async with streamablehttp_client(URL, headers=headers) as (read, write, _):
        async with ClientSession(read, write) as session:
            await session.initialize()
            tools = await session.list_tools()
            print([t.name for t in tools.tools])

            # Read-only: mid prices for every Hyperliquid market
            mids = await session.call_tool("hyperliquid", {"action": "get_all_mids"})
            print(mids.content)

asyncio.run(main())

Tool reference

18 tools. The larger venues (hyperliquid, polymarket and the launchpads) group many operations behind one tool with an action parameter. Actions marked ✍️ change state: they place orders, move funds or sign transactions.

Trading venues

Fund your Hyperliquid account by sending USDC on Arbitrum to the address from get_deposit_address. Withdrawals go back to the same address on Arbitrum.

Polymarket orders settle from your Polymarket deposit wallet, not your main wallet. Use get_deposit_wallet to find it before funding.

Swaps, bridges and transfers

Wallet and market data

Token launches

✍️ launch, simulate_launch, list_launches, get_launch, get_funding, list_creator_launches, get_raffle, get_raffle_claims, list_raffle_stocks, get_readiness, get_keeper_status

✍️ launch, ✍️ buy, ✍️ sell, ✍️ graduate, ✍️ claim_fees, list_launches, get_readiness, list_quote_assets, get_pair_terms, get_pool, get_market, get_candles, get_quote, get_fees

✍️ launch_token, ✍️ buy, ✍️ sell, ✍️ migrate, ✍️ claim_creator_fees, list_tokens, get_pool_state, quote_buy, quote_sell, get_pool_fees, get_creator_fees

list_quote_assets, get_launch_terms, list_tokens, get_claimable_fees, ✍️ enable_holder_fee_sharing, ✍️ sweep_fees, ✍️ claim_fees, ✍️ launch_token (legacy; new launches use openfinance-launchpad-wrapper)

Example prompts

Research

• "Which Hyperliquid perps have the highest funding right now? Show the top 10 with open interest."

• "List every stock perp on Hyperliquid and today's move for NVDA, TSLA and AAPL."

• "What are the odds on Polymarket for the next Fed rate decision? Show me the order book."

• "Who are the top 10 Polymarket traders in sports this month?"

• "Find PEPE on Base and tell me its price, 24h volume and liquidity."

Portfolio

• "What do I hold across all my wallets, in USD?"

• "Show my open Hyperliquid positions with entry price, PnL and liquidation price."

• "What are my open Polymarket positions and my PnL this week?"

Trading (your agent will show you a summary and wait for your yes)

• "Long 0.01 BTC on Hyperliquid at 5x with a limit at 2% below the current price."

• "Buy $25 of YES on the leading Polymarket outcome for the Champions League final."

• "Swap 50 USDC on Polygon for ETH on Base."

• "Swap 1 SOL for BONK."

• "TWAP 500 USDC into ETH perps over 2 hours."

Funding and launches

• "Give me a link to buy $100 of USDC with my card."

• "Launch a token called Rocket (RKT) on Robinhood Chain that burns its trading fees. Simulate it first."

Safety and security

How write actions work. OpenFinance creates your trading wallets when you sign up. When your agent calls a write action (✍️ above), the server signs and submits it from your account. To keep you in control:

• Every write tool tells the agent to confirm with you first. The agent should quote or preview, show you the tokens, amounts, chains, fees and recipient, and wait for your explicit yes in the current conversation.

• Transfers to outside addresses get an extra warning. If the recipient isn't one of your own wallets, the agent must show a clear external-transfer warning before it sends anything.

• Keep your client's approval prompts on. Most clients can ask before each tool call. Leave that on for OpenFinance, and don't add its tools to an auto-approve list. It's your second check if a model ever skips a confirmation.

Protecting your account

• Prefer OAuth sign-in over API keys where your client supports it.

• If you use a key, keep it in an environment variable or a secret prompt. Never commit it, and never paste it into a chat, issue or screenshot.

• Replace your key at once if you think it has been exposed; the old key stops working immediately.

• Revoke clients you no longer use from your OpenFinance account.

• Be careful with prompt injection. Don't ask an agent with trading access to act on instructions from web pages, emails or documents you don't trust, and never let it take an address or amount from that kind of content without checking it yourself.

• You can export your wallets' private keys from Settings → Export Keys in OpenFinance at any time.

This repository is documentation only. It contains no code, no keys, no secrets and no install scripts.

To report a vulnerability, use Report a vulnerability on this repository's Security tab. Please don't open a public issue for security problems. See SECURITY.md.

Pair it with OpenFinance Skills

The MCP server gives your agent the tools. OpenFinance Skills teach it how to use them well: how to size a Hyperliquid perp, read a Polymarket order book, or route a cross-chain swap. They work in Claude Code, Cursor, Codex and most coding agents.

npx skills add openfinance-tech/skills -y

Verify the connection

Check the server is reachable. Without credentials it answers 401 and points your client to sign-in, which is expected:

curl -si https://api.openfinance.tech/agent/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head -3

With your API key it lists all 18 tools:

curl -s https://api.openfinance.tech/agent/mcp \
  -H "x-api-key: $OPENFINANCE_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

You can also browse the tools interactively with the MCP Inspector:

npx @modelcontextprotocol/inspector

Choose Streamable HTTP, enter the URL, and sign in or add the x-api-key header.

Troubleshooting

FAQ

No. OpenFinance is a hosted remote MCP server. You add a URL to your client; there's no package, Docker image or local process.

Any client that supports remote MCP servers over Streamable HTTP, including Claude (web, desktop and Claude Code), ChatGPT, Cursor, VS Code, Codex, Gemini CLI, Windsurf, Cline, Manus, Hermes and OpenClaw. Clients that only support local servers can use mcp-remote.

Every write tool instructs the agent to show a summary and wait for your explicit confirmation, and your client's approval prompts add a second check. Keep both on. Whether an agent follows instructions ultimately depends on the model and client, so stay in the loop for anything that moves money.

Your OpenFinance account. The same wallets are used in the OpenFinance app and over MCP, and you can export their private keys from Settings → Export Keys.

Send crypto to your wallet address (ask your agent for get_wallet_addresses), or ask for a fiat on-ramp link with get_onramp_url. Hyperliquid and Polymarket each have their own deposit address; your agent can fetch both.

The MCP server is the live tool layer: prices, orders, transfers. Skills are playbooks that teach an agent good judgement with those tools. Install both for the best results.

Links

• App: openfinance.tech

• Connect your agent: openfinance.tech/agents

• Docs: docs.openfinance.tech · llms.txt

• Skills: openfinance-tech/skills

• X: @openfinancelabs

• Discord: discord.gg/6EA3z4HjSe

Disclaimer

Trading crypto, perpetual futures, prediction markets and other leveraged or volatile products carries a high risk of loss, and you can lose more than you expect. Nothing in this repository or in the server's responses is financial, investment or legal advice. AI agents make mistakes: review every action before you confirm it. Availability of some markets depends on where you live; you're responsible for following the laws that apply to you.

License

The contents of this repository are released under the MIT License.