Claude for Financial Advisors Consultants: Who Can Set It Up for Your RIA, and When You Need No One

1. The Short Answer: Many RIAs Can Start Alone

If you are looking for a Claude for Financial Advisors consultant, start with a smaller question: which part of the setup can your firm not do alone? For many small RIAs the honest answer is very little.

Anthropic's Claude for Financial Advisors arrived on September 14, 2026. It is a plugin for Claude Cowork, the agentic, task-running mode of the Claude desktop app: eight skills written as plain Markdown and JSON instructions, plus connectors to systems such as Orion, Wealthbox, Addepar and MoneyGuide. The complete guide covers each skill.

Installing it is the easy part, and the /onboarding skill walks each advisor through connecting tools. The slow work happens at the firm level: which plan to buy, how conversations reach your archive, which rules /compliance checks, and who may connect what.

Keep one date in mind while you shop. The plugin launched in September 2026, so nobody has years of experience with it; look instead for years with Claude, with advisor technology or with SEC compliance. Everything here is operational guidance; interpreting the rules stays with your CCO and counsel.

2. When You Need No One

Plenty of firms should set this up themselves. Yours is probably one of them if most of these are true:

• You are a solo advisor or a small team, and one person enjoys setting up software.
• Your stack is short: one CRM (Wealthbox, Salesforce, or Redtail through the Orion connector), one portfolio system with a connector, and Gmail or Outlook.
• Your CCO has already decided how AI conversations and outputs will be kept.
• You plan to start with one or two skills, such as /pre-meeting and /post-meeting.

The do-it-yourself path is well marked. On any paid Claude plan you add the plugin under Customize, then Plugins, or from Anthropic's public GitHub repository (Help Center steps). Each advisor then runs /onboarding, which learns the firm's tech stack, walks through connecting each tool with that vendor's own login, explains the skills, and runs a demo on the Millers, a made-up household, so no real client data comes into it.

A missing connector slows you down without stopping you, because every skill falls back to paste or upload. Anthropic's on-demand launch webinar walks the same path.

If that describes your firm, keep the consulting fee. The step-by-step setup guide covers the rest, and a good consultant will tell you so on the first call.

3. When Outside Help Pays for Itself

Help earns its fee when the work stops being about one advisor and starts being about the firm. Five situations cause that.

• Several systems and several people. Each connector uses your existing vendor account, where access "may require a separate subscription or API key from the respective provider" (Anthropic tutorial), and inherits the signed-in user's permissions. Who connects what becomes an access decision.
• The records question. Claude's audit logs are Enterprise-only, cover 180 days per export and exclude chat titles and content. Capturing content takes the Compliance API feeding your archive, and only the Primary Owner can switch that API on. The /compliance skill says it plainly: "There is no archiving connector yet."
• Your rules. /compliance checks drafts against a checklist it calls "a static snapshot, not a live firm-rules feed." The plugin knows nothing about your disclosures, banned phrases, note format or model names until someone writes them into a private copy (see customizing the plugin).
• The plan. Anthropic recommends Enterprise for RIAs, with a 20-seat minimum; Team costs less but has no audit log export or Compliance API (see small RIAs).
• An exam ahead, or a dual registration. The SEC's fiscal 2026 exam priorities say staff will assess firms' policies for monitoring and supervising their use of AI. Dual registrants add Reg BI and FINRA 2210 to the checks.

None of this is hard for someone who has done it before. It is slow for someone learning it between client meetings.

4. Seven Kinds of Setup Help at a Glance

Seven kinds of help serve RIAs setting up this plugin. Each does part of the job well, and the last column is where the choice gets made.

Few providers publish prices for this work, so ask each for a fixed quote in writing. Many firms combine two rows: one to set up the plugin, and a compliance consultant or counsel for the governance slice.

5. Help You Already Have: Anthropic, Your Custodian, Your Platform

Anthropic. Anthropic sells the licenses and publishes the material: the plugin page, the launch post, the webinar, and Help Center articles on organization plugins, audit logs and the Compliance API. If you buy Enterprise, ask what setup help comes with it.

Read one line of the README before you hire anyone. It calls the plugin a "reference implementation. Not actively maintained or monitored, and not accepting contributions." Once someone adapts it for your firm, you own that copy and its upkeep.

Schwab. Schwab Advisor Services is the only RIA custodian integrated at launch and serves more than 16,000 independent RIAs (Schwab). Firms turn its connector on at the firm level and decide who may use it; clients' dates of birth, Social Security numbers and account numbers are masked; and the cost rides on the firm's Anthropic licenses, with no separate payment to Schwab (Financial Planning).

One wrinkle: the README listed Schwab's connector as coming soon, and until it arrives the skills work from pasted or uploaded data. Schwab can help with the Schwab piece, and only that piece.

Dynasty. Dynasty Financial Partners announced Dynasty AI OS, powered by Claude, for its network of about 60 independent RIAs, with book-level AI chat in early release and a broader rollout planned for the fourth quarter of 2026. If your firm is in that network, ask Dynasty first.

A note on labels. InvestmentNews reported that Anthropic built the product with Dynasty Financial, Schwab and Vanguard, which makes them launch partners. Implementation partner is a looser label, so ask any firm using it what stands behind it.

6. Consultancies: Wealth Tech, Large Firms, Specialists, Compliance

Wealth management technology consultancies and outsourced CTOs. These firms already know RIA stacks: the CRM, the portfolio system, the custodian feeds. F2 Strategy is one example; its outsourced CTO service is "specifically tuned to help small, fast-growing RIAs obtain and integrate the technology they need." If you already have a firm like this, ask it first, including what it has configured in Claude Cowork.

Large advisory firms. PwC has an expanded alliance with Anthropic, and firms of that size suit large institutions running Claude across many teams. An eight-person RIA would be a very small client.

Specialist Claude consultants. A small group of firms focuses on setting up and adapting Claude for investment firms. Their edge is product detail: Cowork, skills, the Enterprise admin settings and MCP connectors (MCP, the Model Context Protocol, is an open standard for connecting Claude to other software). Their limits are small benches, key-person risk, the widest spread in quality of any row, and no legal advice.

Compliance consultancies. ACA Group runs compliance reviews and mock exams and offers advisers AI risk assessments and AI policies, which covers the governance slice. Ask whether a consultancy configures the tool itself or reviews what someone else configured.

For the governance side in depth, see AI governance help for SEC-registered advisers.

7. What a Good Setup Engagement Delivers

Whoever does the work, your own staff included, a finished setup leaves six things in writing. Put this list in front of any provider and ask where each item lands in the proposal.

1. The plan decision. Enterprise or Team, the seat count, spend limits and model defaults, with the reasoning written down.
2. Access. Single sign-on, groups and roles, Cowork switched on, and the plugin pushed to the firm through the organization marketplace, where an Owner can mark it Required. Local MCP servers restricted as your policy requires, since Anthropic warns that a plugin can bundle local MCP servers that run on the user's own machine.
3. Connectors. The systems you use, each with the least access that works and read-only where the vendor offers it, plus a written paste-or-upload fallback for anything without a live connector (see the connectors guide).
4. Your rules inside the skills. A private copy of the plugin that knows your firm: the review standards, disclosures and banned phrases /compliance should apply, the note format /post-meeting should write, and the IPS and model names the rebalance review should use.
5. Records and supervision. On Enterprise, the Compliance API switched on and connected to your archiving vendor; Smarsh, for one, released Smarsh Capture for Claude Enterprise in May 2026, built on that API. Then an AI use policy addendum, supervision steps, and vendor files for Anthropic and each connector, since the amended Regulation S-P expects oversight of service providers that touch customer information.
6. A pilot. A week with two or three advisors working real households, then fixes, then launch, plus a short admin handbook so your team can run it afterward.

Three things usually sit outside a setup quote: the Claude licenses, which you buy from Anthropic; legal advice, which comes from counsel; and firm-wide training, which is its own project (see advisor training).

A proposal that cannot show where all six land is a quote for installation.

8. What a Claude Expert for Financial Advisors Should Show You

Anyone can call themselves a Claude expert. The test is what they can show you on screen and answer without looking it up.

• Walk me through the demo on the fictional Millers household. What will /pre-meeting never do? The answer: send anything to the client, place a trade, or invent plan probabilities or performance.
• Which of our systems have connectors, and what does each need? Expect specifics: which need a vendor subscription or API key, which can write back with approval (Black Diamond can create reports, tasks and rebalance sessions), and the fallback for systems with none.
• Where does /compliance stop? A good answer quotes the skill on its own output, "a draft review for the firm's CCO/compliance officer, not a legal determination or an approval," and knows that anything on performance, hypothetical performance or testimonials is escalated to the CCO.
• How will our conversations reach our archive? Listen for the difference between the audit log, the Compliance API and your archive, and for the five years Rule 204-2 generally requires.
• Where will our firm rules live, and who approves changes? Listen for a private repository synced to the organization marketplace, with your CCO reviewing each new version before it reaches advisors.
• Which guardrails will you leave alone? All of them. The README says some are structural, while others, advisor approval before any write among them, "are enforced by the model following them, not by the runtime," which is a reason to test them and never widen them.
• How many firms have you set this up for? Given the launch date, an honest number is small, so look for nearby work: Claude Cowork rollouts, advisor technology, SEC compliance.

A real expert answers most of these from memory and shows you the rest. General answers point to general expertise.

9. Red Flags in a Setup Proposal

Watch for these in any proposal, including one from a firm you already trust:

• A promise of compliance. A tool cannot make a firm compliant, and the plugin's own /compliance skill never calls a redraft compliant (see the short answer on compliance).
• Licenses resold at a markup, or undisclosed referral fees from the vendors being recommended.
• No archive plan: nothing on the Compliance API, your archiving vendor or five-year retention.
• "We'll build you custom AI" before your advisors have used the eight skills that already exist.
• Fewer clicks through fewer approvals: any plan to drop approval steps or widen what a skill can reach.
• Everything connected on day one, with write access, before a pilot.
• A claim that client data never reaches Anthropic. The records stay at the source, but whatever a connector returns lands in the conversation, which your firm's retention settings then govern.
• Legal conclusions from someone who is not your lawyer.

One of these can be a misunderstanding worth a phone call. Two in the same proposal means keep looking.

10. Where to Start

Before you call anyone, write one page: every system the firm uses, who holds each admin login, how many people will use Claude, and what your CCO wants done with AI records. That page sets most of the scope and makes every quote comparable.

Then take the free route as far as it goes: watch Anthropic's webinar, run /onboarding on one seat with the fictional household, and note where you got stuck. If nothing stopped you, you probably need no one; if records, rules or access did, that is the scope to buy. For timing, see how long setup takes.