Skip to main content
The Legal Talent KYC Connector is a remote Model Context Protocol (MCP) server that lets your team run KYC operations directly from Anthropic’s Claude — Claude.ai, Claude Desktop, Claude Mobile, and Claude Code. Ask in plain language to create verification links, run sanctions/PEP/adverse-media screenings, triage the review queue, and read usage metrics. Each request runs against your own tenant with your permissions.
Your data privacy when using the connector is described in the MCP Connector Privacy Policy. Tool results enter your Claude conversation, so the connector minimizes the personal data it returns.

Prerequisites

  • A Legal Talent account on a plan that includes the Claude connector (connector:connect).
  • At least one verification workflow configured in your dashboard (needed to create session links).
  • A Claude plan that supports connectors (Pro, Max, Team, or Enterprise).

Setup

1

Add the connector in Claude

In Claude go to Settings → Connectors → Add custom connector and paste the server URL:
https://mcp.kyc.legaltalent.ai/mcp
2

Sign in and authorize

Claude opens the Legal Talent sign-in. Log in with your dashboard account and approve the requested scopes on the consent screen. Authentication uses OAuth 2.1 (Authorization Code + PKCE); Claude never sees your password.
3

Start working

Once connected, just ask Claude. It picks the right tools automatically. Try one of the examples below.
Write actions (approving a session, re-issuing a link, running a paid screening) are annotated so Claude asks for your confirmation before executing.

Usage examples

Below are real prompts you can paste into Claude once the connector is connected. The Tools used line shows what runs under the hood.
Session sess_abc123 is about to expire — extend it 7 more days.
Tools used: extend_session

AML / sanctions screening

Run a sanctions, PEP and adverse-media screening on ACME S.A., a company incorporated in Uruguay.
Claude creates the screening and returns a summary (status, risk, hit counts). If your plan has no remaining quota, it returns a secure payment link instead of charging you — it never starts a payment on its own.Tools used: create_screeningget_screening
Screen Maria Gonzalez, born 1985-04-12, nationality UY. Focus adverse media on financial crime.
Tools used: create_screening
Give me the result of screening scr_789 and a link to the PDF report.
Claude returns the summary and a short-lived download link to the finalized PDF report.Tools used: get_screeningfinalize_screening
List the screenings we ran this week and their risk levels.
Tools used: list_screenings

Review queue & compliance decisions

Show the sessions in manual review assigned to me.
Tools used: list_sessions (with status=manual_review, assigned_to=me)
Walk me through session sess_abc123: status, risk score, and any flagged checks.
Tools used: get_session
Approve session sess_abc123 — documents and screening look clean.
Recording a compliance decision is a write action, so Claude confirms before submitting it to the audit trail.Tools used: update_session_status
On session sess_abc123, ask the client to re-upload the ID front — it’s blurry.
Tools used: request_session_corrections
Assign session sess_abc123 to María for review.
Claude resolves María from your team directory, then assigns the session.Tools used: list_team_membersassign_session

Metrics & usage

How are this month’s verification sessions going?
Claude renders an interactive metrics canvas inside the chat — summary, SLA, and daily trend — with filters you can adjust in place. (Falls back to a text summary on hosts that don’t support interactive apps.)Tools used: sessions_metrics_canvas
How much adverse-media screening have we used this month, and how much is left?
Tools used: get_usageget_quota_remaining

Good to know

  • Scoped to your tenant. The connector always acts within the organization you signed in as. It never accepts a tenant chosen by the client.
  • Least privilege. Actions run with a scoped, per-connection API key you can revoke at any time from the dashboard or by removing the connector in Claude.
  • Confirmations on writes. Read actions run freely; anything that changes state (decisions, re-issuing links, paid screenings) is confirmed first.

Support

Questions or issues? Contact support@legaltalent.ai.