Connect your AI agent to Alpaca

Alpaca gives AI agents direct access to trading, market data, and brokerage operations. Pick the tool that fits how you work, connect it in a few minutes, and start in paper trading or sandbox before moving to live.

Choose your path

Start with who you are building for:

Trading for yourself or your business. Use the Trading MCP server or the Alpaca CLI with your Alpaca brokerage account. Both cover the Trading API and Market Data API: account, orders, positions, portfolio history, and market data for stocks, ETFs, options, and crypto.
Building for your end customers or operating as a Broker API partner. Use the Broker MCP server to inspect accounts, review transactions, and build and test operational workflows against your correspondent data.

Then pick how your agent connects:

Tool	Best for	Auth	Get started
Agent plugins	The fastest setup in Cursor, Claude Code, and Codex	Sign in with Alpaca (OAuth)	Install a plugin
Hosted MCP servers	Any other MCP client that supports remote servers	Sign in with Alpaca (OAuth)	Configure hosted MCP
Local Trading MCP server	Running the open-source server yourself, with API keys and toolset filtering	API keys	Trading API MCP
Broker MCP server	Broker API partners	Sign in with Alpaca (OAuth)	Broker API MCP
Alpaca CLI	Terminal, scripts, CI, and focused agent actions with structured JSON output	API keys or `alpaca profile login`	Alpaca CLI
Agent skills	Step-by-step playbooks your agent follows, such as backtesting or paper trading	Uses the Alpaca CLI	Alpaca skills

Not sure? Install the alpaca-trading plugin and connect to paper trading first. You can add the CLI or skills later.

Prerequisites

An Alpaca account. A free paper trading account is enough to start.
An AI client such as Cursor, Claude Code, Codex, VS Code, or another MCP-compatible tool.

Install an agent plugin

Plugins connect Cursor, Claude Code, and Codex to Alpaca's hosted MCP servers. There are no API keys to copy: you sign in with your Alpaca account the first time a server starts.

Plugin	Includes
`alpaca-trading`	Trading MCP (live) and Trading MCP (paper), including Market Data API tools
`alpaca-broker`	Broker MCP (live) and Broker MCP (sandbox)

Cursor

Open the command palette with Cmd + Shift + P (macOS) or Ctrl + Shift + P (Windows and Linux).
Run Open Customize, then click Browse Marketplace.
Click Add Marketplace, choose Import from GitHub, and enter https://github.com/alpacahq/agentic.
Install alpaca-trading, alpaca-broker, or both.
Start an Alpaca MCP server and complete the sign-in in your browser.

Claude Code

claude plugin marketplace add alpacahq/agentic

Start a session and run /plugin. Select the alpaca-plugins marketplace and install alpaca-trading, alpaca-broker, or both. Run /reload-plugins, then use /mcp to start a server and sign in.

Codex

codex plugin marketplace add alpacahq/agentic

Install a plugin from the alpaca-plugins marketplace, then sign in:

codex mcp login <mcp-name>

Configure hosted MCP servers

If you prefer not to use a plugin, or your client is not listed above, add a hosted server to your MCP client configuration directly.

Server	Endpoint
Trading (live)	`https://api.alpaca.markets/mcp`
Trading (paper)	`https://paper-api.alpaca.markets/mcp`
Broker (live)	`https://broker-api.alpaca.markets/mcp`
Broker (sandbox)	`https://broker-api.sandbox.alpaca.markets/mcp`

Each client needs its own sign-in settings. Copy the configuration for your client from the agentic repository. For Broker MCP in VS Code, see Broker API MCP.

Verify the connection

Restart your client if needed, then ask your agent a question for the server you connected. If you get account details back, you are connected.

Trading API: "What is my Alpaca paper account balance and buying power?"
Broker API: "List the five most recently created accounts in my Broker API sandbox."

Try a few more prompts:

Trading API: "Show me AAPL's daily bars for the last week."
Trading API: "What are the most active stocks right now?"
Trading API: "Place a limit order to buy 10 shares of MSFT at $400 in my paper account."
Broker API: "Show recent transactions for account {account_id} in my Broker API sandbox."

Before you use live environments

Start in paper or sandbox. Test agent workflows against paper trading or the Broker sandbox before switching to live.
Review orders. Agents submit orders through Trading API and Broker API; execution is not guaranteed. Review what your agent proposes before you confirm it.
Protect your keys. Never paste API keys into a chat. Keep them in your client configuration or your OS secret store.
Rate limits and data plans apply. Agent traffic counts against your account's API rate limits, and some real-time data requires a market data subscription.

Get help

Request support for another client or report a problem in the agentic repository.
Contact Alpaca Support.

Disclosure

Insights generated by our MCP servers and connected AI agents are for educational and informational purposes only and should not be taken as investment advice. Past performance from models does not guarantee future results. Conduct your own due diligence before making any decisions.