Typing “buy 5 shares of AAPL at market price” into a chat window and watching a real order land in your brokerage account still feels like a magic trick. That’s exactly what Alpaca official MCP server does. It wires Claude directly into Alpaca’s trading and market data APIs, so placing an order, checking a position, or pulling an options chain becomes a sentence instead of a series of clicks.
Table of Contents
- What Is the Alpaca MCP Server, Exactly?
- What Can You Actually Do With It?
- How Do You Set Up the Alpaca MCP Server?
- Is It Safe to Let Claude Place Real Trades?
- Alpaca MCP Server vs. PickMyTrade: Which Fits Your Trading Style?
- Where Is Conversational Trading Headed From Here?
- Frequently Asked Questions
- The Bottom Line
MCP, short for Model Context Protocol, is the open standard Anthropic introduced to let AI assistants call external tools safely. Alpaca built its own MCP server on top of it, and by mid-2026 the broader MCP ecosystem had grown to tens of thousands of servers built on the same idea. This guide walks through what the Alpaca MCP server actually does, how to install it, what it can’t do, and where it fits next to rules-based automation tools like PickMyTrade.
Key Takeaways
- Alpaca’s official MCP server ships with 65 tools covering orders, positions, options, crypto, and market data, and it’s MIT licensed with 900+ GitHub stars.
- Paper trading is on by default, and Alpaca’s own docs warn against ever pasting API keys into a chat window.
- Only 41% of MCP-based agent deployments reach production, and just 8.5% of public MCP servers implement the protocol’s OAuth 2.1 security requirement.
Most coverage of MCP trading tools treats “AI can place trades now” as the whole story. It isn’t. The more interesting split is between conversational MCP servers like Alpaca’s, built for one-off, human-approved trades you type out yourself, and rules-based automation like webhook routing, built to fire the same trade the same way every single time without you in the loop. Confusing the two is how people end up disappointed by one or the other.
If you’ve already read PickMyTrade’s guide to connecting Claude to TradingView, you’ll recognize the theme: that piece points out Claude and MCP still can’t execute a live trade from a TradingView chart alone. Alpaca’s server is the exception that proves the rule, because it connects Claude straight to a real brokerage account instead of a charting platform.
What Is the Alpaca MCP Server, Exactly?
Alpaca’s MCP server is the company’s own official bridge between its Trading API and MCP-compatible AI tools, and it currently ships 65 tools spanning orders, positions, market data, options, and account management. It’s open source, MIT licensed, sitting at over 900 GitHub stars, and was rebuilt from the ground up on FastMCP and Alpaca’s OpenAPI spec for its v2 release.
That “official” label matters more than it sounds. Plenty of trading MCP servers floating around GitHub are community projects duct-taped together over a weekend, with no guarantee they’ll keep working after the next API change. Alpaca maintains this one itself, which is a big part of why it’s become the reference implementation for what a broker-grade MCP server should look like.
Under the hood, MCP just standardizes how an AI model discovers and calls external tools. Instead of Claude guessing at an API’s shape, the MCP server exposes a fixed menu of functions, like place_order or get_positions, with defined inputs and outputs. Claude reads that menu, decides which tool fits your request, and calls it. You never touch code.
What Can You Actually Do With It?
The 65 tools break down into eight functional groups: order management, position management, market data, options trading, crypto trading, account tools, watchlists, and news plus corporate actions. That’s a wider surface than most people expect from a chat-based trading tool.
On the order side, you can place, modify, and cancel market, limit, stop, trailing-stop, and bracket orders across stocks, options, and crypto, all in one conversation. The market data tools return real-time quotes, historical bars, trades, and snapshots, so Claude can pull context before it suggests anything. Options support goes deep too: contract search, multi-leg strategies, and live Greeks and implied volatility data are all in scope, not bolted on as an afterthought.
We tested a handful of commands ourselves during setup. “Show me available option contracts for AAPL expiring next month” returned a filtered chain in seconds. “What’s my current buying power?” pulled account data instantly. The gap between typing a command and seeing Claude’s proposed action was consistently under five seconds, which made second-guessing a suggested trade feel less like friction and more like a built-in pause button.
Watchlists and corporate actions round things out. You can build and manage custom symbol lists, and pull dividend, split, and news data without leaving the conversation. None of these are things you couldn’t already do in Alpaca’s dashboard. What changes is that you’re describing what you want instead of clicking through menus to find it.
How Do You Set Up the Alpaca MCP Server?
Setup takes about ten minutes if you already have Python installed. You’ll need Python 3.10 or newer, the uv package manager, an Alpaca account with API keys (the free paper trading tier works fine to start), and an MCP-compatible client such as Claude Desktop or Claude Code.
Start by creating a paper trading account on Alpaca and generating an API key and secret key from the dashboard. Paper trading uses fake money against real market data, so there’s zero financial risk while you learn the tool. Install uv next if you don’t have it already, since it’s what runs the server via uvx without a manual install step.
From there, add the server to your client’s config. For Claude Desktop, that means editing claude_desktop_config.json to add an alpaca entry under mcpServers, pointing the command at uvx alpaca-mcp-server with your API key and secret key set as environment variables. Claude Code users can skip the JSON entirely and run a single claude mcp add command with the same credentials passed as flags. Restart your client so it picks up the new server. Then ask it something simple, like “what’s today’s market clock?”, to confirm the connection works.
Four environment variables control behavior once it’s running: ALPACA_API_KEY and ALPACA_SECRET_KEY authenticate the connection, ALPACA_PAPER_TRADE toggles between paper and live trading, and ALPACA_TOOLSETS lets you restrict which of the eight tool groups are even exposed to Claude, useful if you only want market data access without giving it order placement at all.
The server also runs on Cursor, VS Code, PyCharm, and Antigravity CLI, plus Claude’s mobile app and ChatGPT if you’re willing to host it remotely instead of running it locally.
Is It Safe to Let Claude Place Real Trades?
Paper trading is the default the moment you install it, and Alpaca’s documentation is direct about the risk once you flip that switch: orders execute against live markets immediately, and you should review every suggestion before approving it. Nothing about the setup nudges you toward going live faster than you’re ready for.
The clearest warning in Alpaca’s own docs is about credential handling: never paste your API keys directly into a chat message. Configure them only in your client’s environment settings or an OS-level secret store. A key typed into a Claude conversation can end up logged or cached somewhere you don’t control, which defeats the point of keeping it out of your code in the first place.
That caution is well placed given the state of MCP security more broadly. Across the MCP ecosystem, only 8.5% of public servers currently implement OAuth 2.1 (the login standard that lets a tool prove its identity without ever seeing your password), and that gap is one reason security researchers keep flagging MCP as an area still catching up to its own adoption curve.
Alpaca MCP Server vs. PickMyTrade: Which Fits Your Trading Style?
Both tools let AI touch your trading account, but they solve different problems. Alpaca’s MCP server is built for conversational, human-approved trades on stocks, options, and crypto inside your own Alpaca account. PickMyTrade is built for systematic, rules-based execution: a TradingView strategy fires an alert, and PickMyTrade routes it straight to a connected futures broker or prop firm account without anyone typing a command in the moment.
That’s not a small distinction. If you want to ask Claude to check your buying power and place a single options trade while you’re watching, Alpaca’s MCP server is the right tool. If you’ve already built a Supertrend or VWAP strategy in TradingView and want it to fire the same way at 3 a.m. without you awake to approve it, that’s exactly the gap PickMyTrade’s TradingView automation closes instead.
We hear a version of this question from traders fairly often: can Claude just run my whole strategy for me? Conversational MCP tools like Alpaca’s are genuinely good at single trades, research, and portfolio checks. They’re not built to sit awake all night watching for a signal. That’s still a job for a dedicated automation layer, not a chat window.
| Alpaca MCP Server | PickMyTrade | |
|---|---|---|
| Asset classes | Stocks, ETFs, options, crypto | Futures, prop firm accounts |
| Trigger | Typed conversational command | TradingView webhook alert |
| Execution style | Human reviews and approves each trade | Automated, rule-based, 24/7 |
| Best for | Research, one-off trades, portfolio checks | Systematic strategies on funded accounts |
| Setup | Python, uv, MCP client config | Webhook URL pasted into a TradingView alert |
Traders who want both aren’t a rare case, either. It’s entirely normal to run Alpaca’s MCP server for hands-on stock and options trading while running PickMyTrade separately for automated futures strategies on an Apex, Topstep, or Tradovate account. Check PickMyTrade’s supported prop firms if that side of your setup is the futures piece you still need to wire up.
Where Is Conversational Trading Headed From Here?
MCP’s growth curve backs up why Alpaca built this in the first place. GitHub search returned nearly 16,000 repositories tagged mcp-server by May 2026, and the number of publicly tracked MCP servers has climbed past 9,000. SDK downloads now run in the tens of millions per month. That’s not a niche standard anymore. It’s the plumbing a growing share of AI tools now assume is there.
That growth pushed MCP from an Anthropic-only project into shared infrastructure. The Linux Foundation’s Agentic AI Foundation now maintains it as neutral ground across vendors, which is a big part of why Alpaca’s server also runs on Cursor, VS Code, and ChatGPT, not just Claude. Expect more brokers to follow Alpaca’s lead over the next year, and expect the production and security numbers above to keep climbing as the standard matures.
Frequently Asked Questions
Yes, the MCP server itself is free and open source under an MIT license. You’ll still need an Alpaca brokerage account, and Alpaca’s own commission structure applies to any live trades you place through it, the same as trading through Alpaca’s regular app or API.
Yes, once you’re comfortable with how it behaves. Setting ALPACA_PAPER_TRADE to false and swapping in your live API keys switches the server from Alpaca’s paper trading environment to real order execution, so treat that flag change as a deliberate decision, not a default setting.
Yes. It’s built on the open MCP standard, so it also runs on Cursor, VS Code, PyCharm, and Antigravity CLI locally, plus Claude’s mobile app and ChatGPT if you host it remotely instead of running it on your own machine.
Yes, and plenty of traders do. Alpaca’s MCP server handles conversational stock, options, and crypto trades on your Alpaca account, while PickMyTrade separately automates TradingView-triggered strategies onto futures brokers and prop firm accounts. They don’t overlap, so there’s no conflict running both at once.
The Bottom Line
Alpaca’s MCP server turns a real brokerage account into something you can talk to, and the 65 tools behind it cover far more than a basic buy-or-sell command. Paper trading by default and a clear warning against pasting API keys into chat make it a reasonably safe way to try conversational trading for the first time.
Just don’t expect it to replace a systematic strategy running while you’re away from your desk. That’s a different job, and it’s the one PickMyTrade was built for. Compare PickMyTrade’s pricing if a futures or prop firm strategy is the piece of your setup that still needs to run itself.
Written by the PickMyTrade Team, covering AI-driven trading tools and TradingView-to-broker automation. Questions about your specific setup? Reach the team through the about page or contact page.
Disclaimer:
This content is for informational purposes only and does not constitute financial, investment, or trading advice. Trading and investing in financial markets involve risk, and it is possible to lose some or all of your capital. Always perform your own research and consult with a licensed financial advisor before making any trading decisions. The mention of any proprietary trading firms, brokers, does not constitute an endorsement or partnership. Ensure you understand all terms, conditions, and compliance requirements of the firms and platforms you use.
Also Checkout: Automate TradingView Indicators with Tradovate Using PickMyTrade
