Jim Nevotti has a line that explains why the broker connectivity problem for AI agents is harder than it looks. "AI that can see your portfolio but not trade it is only half the equation," he said. His company, Connect Trade, just launched an MCP server designed to solve the other half — and the infrastructure angle is what makes this worth covering beyond the press release.
Connect Trade built a unified broker connectivity API that normalizes access to more than 20 brokerage firms across equities, options, and futures. The new MCP server lets AI platforms — including those running Claude, ChatGPT, Gemini, and Perplexity — connect to retail brokerage accounts through a single server-side integration. OAuth authentication means end users connect their accounts with a single click, no API keys required. Real-time WebSocket streaming delivers live data on orders, positions, balances, and options chains — the data layer AI agents need to act with confidence rather than guess.
The architecture is deliberately server-side. Connect Trade is not a consumer app. It is infrastructure that fintech platforms embed. Public.com's Agentic Brokerage product, which launched this week, is one example of a platform that would need exactly this kind of connectivity to function. You cannot build an AI agent that executes trades without a brokerage integration, and you cannot build a brokerage integration at scale without normalizing across 20 different broker APIs, each with their own authentication methods, rate limits, and data formats. That is the problem Connect Trade solved.
The normalized API across all supported brokers is the technical core. When an AI agent queries a user's portfolio through Connect Trade, it receives a consistent data model regardless of which broker the user actually has an account with. The agent does not need to know whether the user is on Schwab or Fidelity — the abstraction layer handles it. That consistency is what makes it possible to build broker-agnostic AI trading workflows, and it is why the OpenAPI specification is designed to be compatible with AI coding tools like Claude, Cursor, and GitHub Copilot, which can generate working broker integrations directly from the spec.
Nevotti frames this as the infrastructure layer the market has been waiting for. The comparison to Stripe is not inappropriate: Stripe did not build the next great e-commerce store, it built the payment infrastructure that made the stores possible. Connect Trade is making the same bet on broker connectivity — that the real value is in being the plumbing between AI agents and brokerage accounts, not in building another trading interface. The 20-broker coverage, the normalized data model, and the OAuth integration are the three things that, in Connect Trade's view, make that bet credible.
The competitive picture is narrower than payments, with a more concentrated set of players. In Connect Trade's positioning, BrokerTec and FactSet build institutional-grade connectivity but not for AI agent use cases. Alpaca and DriveWealth offer API brokerage infrastructure but not at the multi-broker normalization layer the company is claiming. The differentiation, as Connect Trade frames it, is the MCP server — a protocol-level integration that lets any MCP-compatible AI client connect to the broker network without custom code per broker. That framing is the company's own; independent benchmarks comparing the normalization layer against a point-integration approach do not exist, and readers should know that.
The Benzinga Best Embedded Finance award in 2025 is a legitimate signal, though awards are not competitive moats. What matters is whether the normalized broker API and the MCP server can attract fintech platforms faster than the incumbent connectivity providers can add AI-native features. The early access program is the first test of that.
The "half the equation" framing is accurate and useful. AI that reads your portfolio is half. AI that can also act on it is the other half — and Connect Trade just opened that half for business.
