X1 Wealth MCP Server

X1 for your AI assistant

Connect Claude, ChatGPT, or any MCP-capable assistant to your X1 record. Ask your governed record and get answers backed by your own documents and decisions, or an honest "I can't back that" instead of a guess.

Every answer is scoped to exactly what your X1 account can see, and the brain cites the record or declines rather than inventing.

Ask your governed record

Your firm or program record

How your practice or program is learning with X1: what you reviewed, what you accepted or sent back and why, whether you are getting sharper, and what is in your exportable record. Cited, or declined.

  • "How did my practice do with X1's suggestions this quarter?"
  • "What got sent back the most, and why?"

Your own household record

What X1 has on file for your entities, trusts, properties, and policies, what changed recently, and what decisions are recorded. Cited to your documents, or declined. X1 keeps your record, not your balances.

  • "What does X1 have on file for my trust?"
  • "What changed recently?"

Both answer only from your own governed record, with citations, or they decline. Neither invents, and neither reaches beyond what you can see.

Why this is safe to connect

Your access only.

A connected assistant can never see another household's or firm's data, and never more than your X1 account can. An advisor only ever reaches their own assigned clients.

Cited or refused.

The brain answers from your governed record with citations, or it declines rather than guessing.

People stay in charge.

Outside assistants propose and draft; a person confirms what changes a record.

Documents stay in X1.

Raw document contents are blocked on outside connections.

Connect your assistant

OAuth-capable clients

Add X1 as a connector using https://mcp.x1wealth.com, sign in with your X1 account, and approve access. Your assistant then sees only what your X1 account can see. Write access is governed by your X1 role, entitlements, and person-confirmed workflows.

Codex CLI

Codex can connect directly to the MCP resource URL. Add the endpoint, then sign in. X1 grants read access after sign-in and maps write capability from your X1 entitlements.

codex mcp add x1 --url https://mcp.x1wealth.com/mcp
codex mcp login x1

If you edit config.toml by hand, use this entry:

[mcp_servers.x1]
url = "https://mcp.x1wealth.com/mcp"

Direct clients and API keys

In X1 settings, create an API key, and add it to your MCP-capable client as a bearer token. Use this for clients without an OAuth flow.

{
  "mcpServers": {
    "x1-wealth": {
      "url": "https://mcp.x1wealth.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

For Codex without OAuth, keep the key in an environment variable:

export X1_MCP_API_KEY="x1k_..."
codex mcp add x1 --url https://mcp.x1wealth.com/mcp --bearer-token-env-var X1_MCP_API_KEY

A quick browser auth error during setup is expected and safe to dismiss; it does not affect the connection.

Set up your assistant

Once connected, teach your assistant to reach for X1 first. Paste this once into your assistant's custom instructions (Codex, Claude, or ChatGPT), or save it as a skill:

When I mention a client or ask about client work, use the X1 tools first instead of answering from memory. Resolve the client with list_assigned_members, then call get_x1_guide and follow its 'Working a client from Codex or Claude' playbook: pull the client's context and my prior working notes before answering, and at the end of client work save a session note and log decisions back to X1. Every number must come from an X1 tool result and name its source. If X1 declines something, respect the boundary.

The full playbook lives in the connection itself: ask your assistant to call get_x1_guide and it will walk you through working a client, step by step.

More your assistant can do, scoped to your access

A connected assistant can also work with the rest of your X1 account, always limited to exactly what you can access:

  1. 1See your client and household context you already have access to (profile, product state, activity, decision memory, family-office and entity context, meeting prep).
  2. 2Discover shared vault documents by metadata, and search inside your own indexed documents for source snippets while raw document contents stay in X1.
  3. 3Work with your coordination threads: find them, summarize them, and draft replies, briefs, and document requests for you to send.
  4. 4Draft browser evidence missions and check X1-owned mission results without launching a browser from your assistant.
  5. 5Record a decision to your X1 decision record, and draft then confirm a missing-document request, both scoped and reviewed by a person.
  6. 6Save a note, memo, or report to your own X1 account, where it files in your vault with a note that your assistant added it. A professional can also save work for a member, including an uploaded PDF, DOCX, or spreadsheet, as team-only draft work a person reviews in X1 before the member ever sees it.
  7. 7Upload a file (PDF, DOCX, spreadsheet, or image) straight into your own X1 vault, where it is classified and searchable like any other vault document, flagged as added by your assistant.
  8. 8Organize the household entity map from an outside assistant: propose entity cleanup and Suggested ownership for member review, grant explicit entity-stewardship authority, and let authorized stewards add, update, or file vault documents and connected accounts under the right trust, business, property, asset, or Personal bucket after confirmation.

To answer your questions, your assistant can find and quote short passages from your own documents, always scoped to what you can access and recorded in X1. Full documents and downloads stay in X1. Anything that changes a record is proposed or drafted for a person to confirm.