How to Connect Public to an AI Assistant

Connecting Public to an AI assistant means giving an assistant like ChatGPT or Claude permission to see your real Public holdings, so its answers are about your actual portfolio instead of hypothetical examples. The AI does not log into Public directly. Instead, a secure aggregator sits in the middle: you authenticate with Public through Public's own flow, the aggregator receives scoped access, and the AI tool reads your positions through the aggregator.

Public is a US broker known for fractional investing and a social, community-driven angle, and it supports read plus trade-if-available access. That makes Public a strong fit for AI-driven analysis, and also one of the brokers where an assistant can go a step further and prepare a trade. The trade-if-available part is gated: trading is a separate permission from reading, and you approve every order before it executes. So Public covers both ends of the spectrum, analysis first, approved execution second, unlike read-only brokers such as Robinhood, where outside connections cannot place trades at all.

There are three practical ways to get your Public holdings in front of an AI assistant, ordered here from most capable to least.

• 1. An MCP connector backed by an aggregator. A Model Context Protocol connector exposes your Public account to Claude Desktop, ChatGPT, or Cursor, so the assistant itself can read your positions live and, because Public supports trading, prepare orders you approve. The connector links your Public account securely. Walnut offers a portfolio-aware MCP connector that does exactly this. Best if you already work inside an AI client.
• 2. A purpose-built app that connects your broker for you. A hosted AI investing app connects your Public account securely and adds an assistant you talk to, with no code. You log in at Public, the app reads your holdings, and you ask questions through Claude, ChatGPT, or a built-in assistant. Best for most people. Walnut works this way too.
• 3. Manual copy-paste. You paste a list of your Public holdings into ChatGPT or Claude by hand. It works for a one-time snapshot, but prices go stale immediately, you have to keep re-pasting, the AI cannot place any trades this way, and long lists invite arithmetic mistakes. Fine for a quick look, weak for ongoing analysis.

For a deeper comparison of connector options, see the best MCP connectors for brokerages. To see which brokers support analysis versus approved trading, see which brokers have an AI assistant.

Walnut connects your Public account securely, so your Public login is never stored by Walnut. The flow takes a few clicks:

1. Create a Walnut account and open the brokerage-connection screen.
2. Pick Public from the list of supported brokers.
3. Log in at Public. Walnut opens Public's own authentication flow, so your password is entered at Public and never seen by Walnut.
4. Authorize access. The connection defaults to read-only for analysis; because Public supports trade-if-available, you can also authorize approved trading.
5. Start chatting. Ask about your holdings, performance versus the S&P 500, or concentration, through Claude, ChatGPT, or Walnut's built-in assistant.
6. Approve any trade. If you ask the assistant to place orders, it prepares them and you confirm each one. Public enforces a $5 minimum per fractional order, so each leg must be at least $5.
7. Revoke any time. Disconnect Public from Walnut's connections screen, or revoke the authorization from your Public security settings, and the access stops.

Because Public supports both reading and approved trading, Walnut treats a connected Public account as a full analysis-and-execution source. You can keep it read-only and use Public purely for tracking, or enable trading and let the assistant prepare orders you approve. If you prefer to work inside your own AI client, you can use Walnut from there by adding its MCP server URL, https://walnutinvest.com/api/connector/mcp, as a custom connector in Claude, ChatGPT, or Codex, then signing in with a free Walnut account and approving read-only access (Walnut never places trades on its own; any order stays with the broker and your approval). The full per-client setup lives on the Walnut MCP connector page.

Once Public is connected, the assistant can answer questions about your real positions instead of hypothetical ones. Useful prompts include:

• “How is my Public portfolio doing versus the S&P 500?” Each holding framed against a benchmark, as a window return rather than realized profit and loss (broker feeds rarely pass cost basis).
• “Where am I most concentrated?” Which positions dominate the account and how lopsided the mix is.
• “Which of my holdings are lagging?” Relative performance across your positions, surfaced in plain language.
• “What themes does my Public account lean into?” Whether your holdings cluster around AI, semiconductors, or another theme.
• “Prepare orders to add to this portfolio.” Because Public supports trading, the assistant can prepare orders you then approve, with each leg at least $5.

The assistant reads and analyzes, and on Public it can also prepare orders you approve. You can ask about a specific stock or a theme you want exposure to, framed against your real holdings.

This is where Public differs from read-only brokers. The Public connection supports read plus trade-if-available. Reading is the default: the assistant sees your positions, performance, and concentration. Trading is a separate permission, and even when enabled, nothing executes automatically. The assistant prepares an order, you review it, and you approve it before it goes to Public.

One concrete detail to plan around: Public enforces a $5 minimum per fractional order. So if an assistant prepares orders across a multi-stock portfolio, each individual leg has to be at least $5. For a five-stock portfolio where the smallest target weight is 10%, that means roughly $50 or more total. This is a Public platform rule, not something an AI tool can override. Contrast this with connecting Robinhood to an AI assistant, where third-party connections are read-only and the assistant cannot place trades at all. Public is one of the brokers, alongside Alpaca, Schwab, Tradier, and Webull, where approved AI-assisted trading is possible.

Public is one of the few brokers that already ships its own built-in AI assistant, Alpha, so a fair question is why you would connect Public to an outside assistant at all. The short answer: Alpha and an external connection solve different problems, and many Public users end up using both. Verify current Alpha capabilities in the Public app, since first-party AI features change often.

When Public's Alpha is the right tool. Alpha lives inside the Public app and is built around your Public account, so it is the fastest way to ask context-aware questions while you are already in Public: what a position is, why a holding moved, what an earnings print means. It is first-party, so there is nothing to connect and no aggregator in the middle. If everything you own is at Public and you just want quick, in-app answers, Alpha is the natural starting point.

When connecting Public to Claude or ChatGPT is the right tool. A built-in assistant only knows the app it lives in. If you want to use a specific model you already trust (Claude or ChatGPT), keep the analysis in the same assistant you use for everything else, combine Public with accounts held at other brokers, or work inside Claude Desktop through an MCP connector, you connect Public outward through a secure aggregator. That is also the path when you want a portfolio framed consistently against a benchmark like the S&P 500, thematic portfolios built around a thesis, or approved order preparation that spans more than what a first-party tool exposes. Walnut is the AI investing assistant that talks to the broker you already have and places the trades you approve. It connects your Public account securely, so you can analyze your real holdings through Claude or ChatGPT and, because Public supports trade-if-available, prepare orders you approve. Walnut is not an investment adviser.

The two are not mutually exclusive: use Alpha for quick in-app lookups, and an outside connection when you want a particular model, a cross-broker view, or an assistant that stays with you outside the Public app. For the wider field of first-party broker AI, see which brokers have an AI assistant.

It can be safe, and the security model rests on three things. First, credential handling: through a regulated aggregator, your Public login is entered at Public's own authentication flow and is never stored by the AI tool. Second, scope: the connection defaults to read-only, and trading is a separate permission you grant deliberately, with every order requiring your approval. Third, control: you can revoke the connection at any time, from the app's connections screen or from your Public account security settings.

The pattern to avoid is any tool that asks for your Public password directly rather than routing you through the broker's own login. The Model Context Protocol, the open standard introduced by Anthropic in late 2024, was designed around scoped, permissioned access, so reading and trading are separate grants. Confirm the tool uses a regulated aggregator, confirm you control whether trading is enabled, and confirm you can revoke it before you connect. For a fuller treatment, see is it safe to connect your brokerage to an AI.

The connectors and tools below all get your Public holdings in front of an AI, but they differ in what they do and whether they can place approved trades on Public.

To connect Public to an AI assistant: pick a tool that uses a regulated aggregator, create an account, choose Public from the broker list, log in through Public's own flow, and authorize access. The connection starts read-only for analysis; because Public supports trade-if-available, you can also enable approved trading, with each order confirmed by you and each fractional leg at least $5. From there you ask Claude, ChatGPT, or a built-in assistant about your real holdings. With Walnut, the same connection also works as a portfolio-aware MCP connector if you prefer to stay inside Claude Desktop.

You connect Public to an AI assistant through a secure connection, which exposes your real holdings to ChatGPT or Claude so the assistant's answers are about your actual portfolio. What sets Public apart from read-only brokers: it supports trade-if-available, so beyond analysis an assistant can prepare orders you approve, with a $5 minimum per fractional order and your confirmation on every trade. Walnut, an AI financial assistant that knows your portfolio, connects your Public account securely and lets you analyze your real holdings through Claude or ChatGPT, with the connection read-only by default and revocable at any time. Walnut is not an investment adviser.

The connection is to Walnut, and the assistant then reads Walnut, so the same setup works in every client that supports remote MCP servers rather than needing a separate integration each time. The server URL is:

https://walnutinvest.com/api/connector/mcp

Verified today: Claude (Desktop, claude.ai and Claude Code), ChatGPT (Developer Mode on a paid plan), Codex and the Codex CLI, and Cursor. ForPerplexity, Gemini, Grok and Copilotwe do not verify custom connector support, so that page explains how to check yours in under a minute rather than giving steps that may not apply.

Step-by-step:ClaudeorChatGPT.

Public accounts tend to mix individual companies with themed holdings, which makes sector clustering the first thing to check. Every question below is answered against the positions in the account rather than from general knowledge.

Portfolio analysis

• “How has my Public account done against the S&P 500 this year?”
• “What share of it sits in my single largest holding?”

Research

• “What does my largest holding actually do, and why might it have moved this week?”

Cost

• “What is my weighted average expense ratio across this account?”

Rebalancing

• “How far has each position drifted from its target weight?”

Planning a purchase

• “If I added $500 to this idea, would I be more concentrated or less?”

More inwhat to ask once it can see your portfolio. If you hold accounts elsewhere too, connect those as well: seeRobinhood, E*TRADE or Webull.