Waitlist2026 State of AI at Enterprise RIAs2026 State of AI at RIAsOut Nov 6On demandOther eventsNext: ADVISE AI, Oct 20–22Booth 204

BlogGuide

Claude Code can't connect to Tamarac

Keep Claude Code and add Caddi: Caddi finds the repetitive reporting and reconciliation in Tamarac, automates the high-volume steps through the API, hands the judgment calls to Claude as a skill, and monitors both.

Tamarac (Envestnet) is the portfolio-management, reporting, and rebalancing platform for RIAs.

  • 99%of agent runs complete successfully
  • ~45 minon average to build your first agent
  • 1M+back-office tasks completed

01Reaching it

Claude Code is a coding agent, not a connector.

It reads and writes files and runs commands in a terminal. Reaching Tamarac means writing custom API code yourself, where an API exists at all.

Claude Code

Custom integration code your team writes, with tokens and secrets in scripts.

Caddi

Connects to Tamarac through its API, with scoped access the firm controls and every run logged.

02Running it

A working prototype is not a production run.

Claude Code can sketch the steps under a developer's eye. Operations and advisory teams pull positions and performance, build client reports, and reconcile Tamarac against the CRM and custodian. In production the same input has to give the same result, hundreds of times a week.

Claude Code

Output can vary run to run, so someone supervises every run.

Caddi

Runs the high-volume steps as code across Tamarac, Salesforce and Microsoft Outlook, the same way every run, and flags anything unsure for review. Judgment-heavy work goes to Claude as a skill.

03Governing it

Nobody can see what ran, or where.

An API change, a new login screen or an expired token silently breaks the script, and nothing tells the firm what ran, what failed, or what is still done by hand before the reporting and reconciliation stops.

Claude Code

Tokens, scopes and secrets handled in ad-hoc scripts, outside firm controls.

Caddi

Caddi Govern monitors every Caddi run, and the work Caddi hands to Claude Code reports each decision and API call back to it, in one firm-wide view: what is automated, where it runs, and what is still manual. Caddi is SOC 2 attested.

Caddi

See how Caddi AI Agents can run the high-volume work inside Tamarac and hand the judgment calls to Claude

Or sign up free

Frequently asked questions

Can Claude Code connect to Tamarac?

Not in a governed, production-ready way. Claude Code is a coding agent, reaching Tamarac means writing and maintaining custom API integration code yourself, with credentials handled outside your firm's controls. You don't have to replace Claude Code to fix it: Caddi connects to Tamarac directly, runs the high-volume work there, and hands judgment-heavy work to Claude as a skill.

Do I have to choose between Claude Code and Caddi?

No. Caddi Discover finds and ranks the repetitive work across Tamarac and your other systems, with evidence. Each finding is automated in Caddi or exported to Claude as a skill, whichever is the better tool: Caddi for high-volume, high-accuracy work that has to run the same way every time, Claude Code for individual, judgment-heavy work. Caddi Govern monitors every Caddi run and the work exported to Claude Code, which reports each decision and API call back, in one firm-wide view.

How does Caddi connect to Tamarac?

Through a real Tamarac connection. Caddi reads and writes Tamarac as part of an automated workflow, with scoped, permissioned access and a run-by-run audit trail (Caddi is SOC 2 attested).

Do I need a developer to set this up?

No. You show Caddi the workflow over a screen share and Caddi builds and maintains the automation for you, including the Tamarac connection. There's no integration code for your team to write or keep running.

Is my client data safe?

Yes. Caddi uses scoped, permissioned access and runs production work as deterministic code with a full audit trail. AI is used only at setup to understand the workflow, not to make decisions on every run, and Caddi is SOC 2 attested.