Anthropic’s Claude for Financial Advisors Ships Connectors, Not a Model — and the Connector List Is the Product

On September 14, 2026, Anthropic released a plugin for Enterprise customers that maps every system of record in a wealth-management practice — and the map is what it actually sold.

What Happened

Claude for Financial Advisors is a plugin, not a new model. It is delivered to Enterprise customers through Anthropic’s Cowork plugin browser and consists of two things: connectors that reach into the systems where adviser data already lives, and workflow skills built around how an adviser’s day actually runs. Before this release, Cowork already held seven general connectors — Microsoft 365, Salesforce, DocuSign, Box, FactSet, S&P Global, and Morningstar. The new release adds eleven wealth-specific connectors on top of those, bringing the total integration surface to eighteen.

The eleven new connectors are: Charles Schwab (custodial balances, positions, transactions, cost basis, alerts, money movement); BlackRock (model portfolios, Advisor Center analytics); Vanguard (model portfolios, advisor investment solutions); Addepar (public and private markets portfolio data and analytics); iCapital (alternative holdings, NAV, commitments); Envestnet (Tamarac accounts, MoneyGuide financial plans); Orion (portfolio reporting, Redtail CRM insights); SS&C Black Diamond (portfolio, performance, holdings, rebalancing); Wealthbox (client records, meeting history); Wealth.com (estate and tax position data); and Zocks (AI meeting assistant, client intelligence). Eight named workflow skills run across that surface: onboarding, alternative-investments brief, estate-and-tax brief, portfolio-rebalance review, post-meeting, pre-meeting preparation, prospect intake, and compliance and AI-policy.

The same day, Schwab Advisor Services said it is working with Anthropic to bring the product to the 16,000-plus independent registered investment advisers it serves through its custodian platform. That is an announced collaboration, not a completed deployment, and no uptake figure was attached to it. Josh Brown, chief executive of Ritholtz Wealth Management, offered the sharpest practitioner read of the launch.

The Structural Read

Walk the eleven names by function and the product design becomes legible. Addepar, Envestnet, Orion, and SS&C Black Diamond are where portfolios get reported and reconciled — they are the systems a back office lives in. Schwab and Vanguard are where assets sit in custody. BlackRock and iCapital are where product originates. Wealthbox and Zocks hold the client relationship: records, meeting history, captured intelligence. Wealth.com holds the estate documents. Assembled, that is not a feature set. It is a map of a wealth-management practice’s systems of record, and assembling it — getting commercial agreements, defining payloads, building authenticated connectors — is the part that was hard.

This is the Business Engineer framework called the Map of AI applied at the vertical layer. Anthropic did not ship a better model for advisers. It shipped a permission surface: the right to sit on top of somebody else’s data, authenticated and structured, in eighteen systems simultaneously. That surface is where defensibility is being built in vertical AI — not in model weights, which commoditise, but in the integration layer and the permission to occupy it. Any sufficiently capable model can draft prose. Only the model with authenticated connectors into Addepar and iCapital and Wealth.com can see the portfolio, the alternative allocation, and the estate plan in the same context window without a human doing the paste job.

The Schwab line is the real distribution event, and it reflects how this industry actually moves. Reaching more than 16,000 independent registered investment advisers through a single custodian relationship is a structurally different motion from selling seats to 16,000 individual firms. In wealth management the custodian is the chokepoint: it holds the assets, it is already wired into the adviser’s daily workflow, and its technology recommendations carry unusual weight precisely because switching custodians is expensive and rare. Distribution through that relationship reaches practices that would never run an AI procurement process on their own initiative. The qualification is important: Schwab Advisor Services said it is working with Anthropic to bring the product to those advisers — an announced collaboration, not a deployment, with no adoption or uptake figure yet attached. What can be assessed is that the route to market was chosen as deliberately as the product. In this sector, the route is usually the harder problem.

There is a useful architectural contrast running in parallel this week. The same question — how do you apply frontier AI capability to confidential professional material in a regulated industry — produced two opposite answers in the same fortnight. Latham & Watkins bought NVIDIA H200s and brought open weights inside its own perimeter. The weights travel to the data because the obligation is that the corpus must not leave. Anthropic’s answer for financial advisers runs the other way: the model stays where it is and reaches into the systems of record through connectors. The model travels to the data’s systems. This piece takes no view on which architecture is better. The point is that the constraint selected the architecture. Where the professional obligation is that material must not leave a perimeter, a firm ends up owning silicon. Where the obligation is that activity must be governed, supervised, and recorded, a connected model with permissions and logging may serve the purpose differently. Which fits depends on the shape of the duty, not on the technology.

The compliance and AI-policy skill deserves more attention than it will get in the initial coverage. Most product launches into regulated industries wave at compliance in general terms. This one names the rule. The skill screens client-facing language against the SEC Marketing Rule, flags regulated content, documents review workflows, and includes a self-guided workflow for documenting a firm’s use of AI under SEC rules. Anthropic separately recommends Enterprise plans for registered investment advisers because they include the audit logs that support recordkeeping requirements. That addresses what was actually blocking adoption in this category. The constraint was never that AI couldn’t draft adequate client-facing prose. The constraint was that supervised communications must be reviewable and the review process must be documentable — which means a tool that generates client-facing language is operationally useless unless the review trail comes with it. Naming a specific rule is also a claim a compliance officer can test against their own obligations, which is a different posture from a general assurance about safety or care.

That said, it should not be overread. Nothing in this release establishes that the screening is accurate, that any firm’s obligations are thereby satisfied, or that any regulator has reviewed or endorsed the implementation. The items that remain genuinely undisclosed are narrower than they first appear — pricing, whether adviser or client data is used for model training, and where processing occurs — but for a product whose named skills include compliance, those are precisely the specifics on which serious buyers in this industry will decide.

Three Implications