Getting StartedDeepvue MCP

Deepvue MCP — Run Verifications From Your AI Tools

Connect Claude, Cursor, or any MCP client to the Deepvue MCP server. Run PAN, KYB, court record, FIR, PEP/AML, and DigiLocker verifications using natural language, authenticated with your Deepvue Client ID and Client Secret.

Overview

The Deepvue MCP server lets any MCP-compatible AI tool — Claude Desktop, Claude Code, Cursor, and others — run real Deepvue verifications through natural language. Instead of writing HTTP requests, you ask your AI assistant to "verify this PAN" or "screen this person for court records," and it calls the corresponding Deepvue API on your behalf.

Each customer connects with their own Client ID and Client Secret, so every call runs against your account, consumes your wallet balance, and respects your plan's enabled products and rate limits.

This is different from the Docs Search MCP. The Docs Search MCP (docs.deepvue.ai/_mcp) only searches this documentation. The Deepvue MCP described here (mcp.deepvue.ai/mcp) executes live verifications against your Deepvue account.

Server endpoint

https://mcp.deepvue.ai/mcp

This is a remote, OAuth-authenticated MCP server using streamable HTTP transport.

Prerequisites

Have an active Deepvue account

Sign up at the Deepvue Dashboard and ensure your wallet has balance for the products you want to use.

Get your credentials

In the dashboard, open the Credentials tab and copy your client_id and client_secret. These are the same credentials used for direct API authentication.

Install an MCP client

Use Claude Desktop, Claude Code, Cursor, or any other MCP-compatible AI tool.

Your client_secret authorizes real, billable verifications. Treat it like a password — never share it or commit it to version control.

Connect the server

The recommended method for Pro, Team, and Enterprise plans.

  1. Open Claude Desktop, then go to Settings (Claude menu → Settings, or ⌘,) and select the Connectors tab.
  2. Scroll down and click Add custom connector. Set Name to Deepvue and URL to https://mcp.deepvue.ai/mcp. Optionally, if you have your Client ID and Client Secret handy, enter them here now — you won't be asked for them again in the next step. Then click Add.
  3. The connector appears in the list — click Connect. Your browser opens the Deepvue login page. If you didn't pre-fill them, enter your Client ID and Client Secret, then click Authorize.
  4. Back in Claude Desktop the connector shows Connected and its tools are registered. In any chat, open the tools/connector menu (the ⚙ / search icon near the message box) and make sure Deepvue is enabled.

Available tools

Once connected, your AI client registers the Deepvue verification tools. The exact tools you see depend on the products enabled on your plan.

ToolWhat it doesMode
identity_verifyVerify an Indian PAN against an authoritative PAN/tax-registry source. Returns the holder profile: name, address, DOB, gender, masked Aadhaar, and Aadhaar linkage status.Synchronous (slow, ~2-5s)
entity_verifyLook up an Indian business or director by gstin, cin, din, or pan. Routes to GSTIN Advanced (filings, promoters, jurisdiction), company-registry CIN (master data, directors, charges), DIN (director profile), or the GSTINs linked to a PAN. For an individual's PAN identity, use identity_verify instead.Synchronous (~1-5s)
court_record_checkScreen an individual against Indian court records (civil, criminal, and tribunal cases). Each call is a new billable screening; if results exceed the ~90s budget it returns completed: false with a transaction_id for court_record_fetch.Async (blocking, ~90s)
court_record_fetchRecovery companion — fetch a court-record result by transaction_id if the original check did not finish in time.Async fetch
fir_record_checkScreen an individual against FIR (First Information Report) police records by name and address. Each call is a new billable screening; if results exceed the ~90s budget it returns completed: false with a transaction_id for fir_record_fetch.Async (blocking, ~90s)
fir_record_fetchRecovery companion — fetch an FIR result by transaction_id.Async fetch
pep_sanction_aml_checkScreen an individual against PEP (Politically Exposed Persons), sanctions, and AML watchlists by name. Each call is a new billable screening; if results exceed the ~90s budget it returns completed: false with a transaction_id for pep_sanction_aml_fetch.Async (blocking, ~90s)
pep_sanction_aml_fetchRecovery companion — fetch a PEP/sanctions/AML result by transaction_id.Async fetch
digilocker_consent_initiateStart a DigiLocker KYC consent flow — issues a link the user authorises in their browser. Returns a redirect_url and consent_id; show the URL to the user, then call digilocker_consent_fetch with the consent_id to complete.Synchronous (~2s)
digilocker_consent_fetchBlock until the DigiLocker run started by digilocker_consent_initiate completes (~90s budget). Returns granted with the verified KYC payload (name, DOB, gender, mobile, address, masked ID, DigiLocker ID), or failed, expired, or pending (still running — call again).Blocking (~90s)

The _fetch tools are recovery companions for the asynchronous flows. A court_record_check, fir_record_check, or pep_sanction_aml_check call submits the request and waits in-line for up to ~90 seconds. If results aren't ready in time, it returns a transaction_id — pass that to the matching _fetch tool to retrieve the result later.

Callers are responsible for having a lawful basis and a legitimate purpose to run these checks. Only the DigiLocker flow collects the individual's explicit consent — all other checks run server-side without one.

How verifications run

Synchronous tools

identity_verify and entity_verify return in a few seconds. Just ask, and the result comes back in the same turn.

Asynchronous screening

court_record_check, fir_record_check, and pep_sanction_aml_check submit a request and block for up to ~90 seconds while court records, police records, or watchlists are queried. If the upstream source is slow, you'll get completed: false with a transaction_id — ask your AI tool to fetch it again, which calls court_record_fetch, fir_record_fetch, or pep_sanction_aml_fetch.

digilocker_consent_initiate returns a redirect_url that the end user must open in a browser to authorize the Aadhaar share via DigiLocker. Your AI tool then calls digilocker_consent_fetch to retrieve the verified KYC payload once consent is granted.

DigiLocker verification requires explicit end-user consent. Only initiate the flow for users who have agreed to share their Aadhaar details, and present the redirect_url to that user — not to a third party.

Example prompts

Once Deepvue is enabled in your chat, try prompts like:

  • "Verify PAN ABCDE1234F and show me the holder's name and Aadhaar linkage status."
  • "Look up GSTIN 27AAACR5055K1Z5 and list the promoters and filing status."
  • "Pull the MCA company master data and directors for CIN L17110MH1973PLC019786."
  • "Screen Rahul Sharma at 42 MG Road, Bengaluru for any pending criminal court records."
  • "Check Rahul Sharma against PEP, sanctions, and AML watchlists."
  • "Start a DigiLocker Aadhaar verification and give me the link to share with the customer."

The AI tool selects the right Deepvue tool, runs the verification against your account, and returns a grounded, plain-language answer.

Troubleshooting

Security and billing

  • Every call runs against your Deepvue account and consumes wallet balance — monitor usage in the Dashboard.
  • Authorize the connector only on machines you trust; the OAuth session grants verification access until it expires.
  • Follow the same security best practices you use for direct API access.
  • For data handling and consent obligations, review Consent and Compliance.