Connect Claude, ChatGPT, or your own agent

Use this when

you live in Claude or ChatGPT and want to search deals, publish mandates, or build lists from a conversation.

Connect your AI assistant to PrivSource to search deals, manage mandates, and build buyer and target lists from a conversation. The connection uses your existing account and plan access, and works on free and paid buyer and seller accounts.

This guide walks you through enabling access and connecting Claude. Other supported MCP clients use the same PrivSource server URL. For an overview of everything you can do, see PrivSource MCP for Claude at privsource.com/mcp-for-claude.

Step 1: Enable access in PrivSource

In the left navigation, click Connect to Claude. If you see Enable MCP Access, click it.

Connect to Claude with Enable MCP Access marked

Step 2: Add PrivSource to your assistant

In Claude, open Customize, then Connectors. Select the plus button, then Add custom connector.

Name the connector PrivSource and enter this URL:

https://api.privsource.com/mcp

Leave the advanced OAuth fields blank and select Add, then Connect.

For another MCP client, follow its custom connector setup using the same URL. See our developer docs at privsource.com/docs for connection details, or the step-by-step setup guide at privsource.com/docs/claude-setup.

ChatGPT, Copilot, and other assistants

For ChatGPT, add a custom MCP server from Plugins where your plan and workspace allow it. Use the same server URL and complete PrivSource authorization. If you cannot create a connection, ask your workspace administrator. See OpenAI’s current setup instructions.

For Microsoft Copilot Studio, open the Copilot instructions on PrivSource’s connection page. Create a Model Context Protocol tool with OAuth 2.0 and Dynamic discovery, then add it to your agent and authorize your team.

For current Claude menu details, see Claude’s custom connector guide.

Assistant menus and availability can change. The connection page includes illustrated Claude and Copilot instructions.

Step 3: Authorize your team

Follow the PrivSource sign-in prompt. Choose the team you want to connect, then approve access.

You can inspect a connection from Connect to Claude in PrivSource. Select View beside the client to review or revoke it.

Step 4: Try a task

Ask something specific, such as:

"Using PrivSource, find live HVAC deals in Texas with $1M+ EBITDA."

The assistant uses PrivSource tools within your team's existing access.

What you can do from Claude

Buyers

Sellers

  • Review the buyers matched to your deals and choose who sees each one

  • List a deal from a teaser, CIM, or description

  • Build a buyer list, or expand one you already have with net-new buyers

Each workflow has a copy/paste prompt and a step-by-step guide at privsource.com/mcp-for-claude. Relevant help articles also include a Do this with Claude section.

If the connection stops working

Remove the connector in your assistant and reconnect it. If it still shows Pending or fails to respond, contact support@privsource.com .

Note: Connecting an assistant doesn't add a membership or unlock tools outside your plan.

Frequently asked questions

Does the connection cost extra?

No. The connection is included with your account. Tools follow your existing plan access.

Does it work on a free account?

Yes. Buyer and seller teams can connect on any plan. Your assistant sees the same tools you see in PrivSource.

Do this with Claude

Connect PrivSource to Claude once and run this workflow from a conversation. Same account, same access.

Agent instructions

The skill file PrivSource publishes for this workflow. Give it to any agent that supports skills, or read it to see exactly what the tools do.

Raw SKILL.md
Show the full instructions

PrivSource MCP Connector

PrivSource MCP exposes account-scoped tools for live deals, seller deal submission, buyer- and seller-side mandate matches, buyer lists, target lists, and Contact Agent enrichment.

Connect

  • MCP endpoint: https://api.privsource.com/mcp
  • Auth: OAuth 2.0 authorization code with PKCE.
  • Scope: mcp:api.
  • Users authorize a PrivSource team. Tool availability depends on that team's buyer/seller accounts and feature flags.
  • Human setup docs: https://privsource.com/docs/mcp-integration.

First Call

Call privsource_whoami before choosing a workflow when account scope is unclear.

Use the returned apis:

  • deals: call privsource_whoami first. Use get_deal for all buyers. Use search_deals only when the account has All Deals access. Otherwise use list_buyside_mandates and get_buyside_mandate_matches.
  • pending-deals: use create_pending_deal to submit a new seller deal for PrivSource admin review and get_pending_deal to track its status through to a live listing.
  • seller-deals: use update_seller_deal to edit deals, archive_seller_deal to archive, and restore_seller_deal to restore. Use set_seller_deal_auto_share to set automatic sharing.
  • buyside-mandates: use buyside mandate tools to manage the buyer account's own published acquisition criteria and review the deals PrivSource matched to them.
  • seller-mandate-matches: use seller mandate match tools to review the buyer mandates matched to the seller account's live deals and contact those buyers.
  • buyer-lists: use buyer-list tools for seller-side acquirer research.
  • target-lists: use target-list tools to build lists of target companies that match filters.
  • contact-agents: use Contact Agent tools for standalone contact enrichment.
  • buyer-profile: read and edit the buyer firm profile.
  • credits: read the credit balance and recent activity.

Tool Selection

  • Buyer searching for acquisition opportunities: use deal tools.
  • Buyer creating or maintaining their own acquisition mandate for sellers to find: use buyside mandate tools (create_buyside_mandate, update_buyside_mandate, publish_buyside_mandate, ...).
  • Buyer reviewing the deals PrivSource matched to a mandate, or tracking saved deals through Pipeline stages: use get_buyside_mandate_matches and update_buyside_mandate_match. Use request_buyside_mandate_match_follow_up only when the buyer asks PrivSource to follow up with an advisor.
  • Seller reviewing buyer mandates matched to a live deal: start with list_seller_deals, then use get_seller_deal_matches, get_seller_mandate_match, and update_seller_mandate_match. Use share_seller_deals_with_buyer only when the user asks to contact the buyer and confirms which matched deals to share.
  • Seller editing a pending submission: use update_pending_deal. For deletion, return the link from get_pending_deal_deletion_link.
  • Seller managing a deal: use update_seller_deal to edit content, archive_seller_deal to archive, or restore_seller_deal to restore. Find archived deals with list_seller_deals and status: archived. Use set_seller_deal_auto_share for automatic sharing.
  • Seller submitting a new deal to be listed on PrivSource: use create_pending_deal, then get_pending_deal to track review status.
  • Seller building a list of potential acquirers for a company being sold: use buyer-list tools.
  • Buyer building a list of target companies that match filters (sector, size, geography): use target-list tools (create_target_list, refine_target_list_filters, get_target_list_results, ...).
  • User has a set of companies or named contacts and wants emails/phones: use Contact Agent tools. When the user has both a buyer and a seller account, pass account_type to choose the account that pays. Each request takes up to 100 contact rows or 500 email-check rows; send more requests for more rows. Rename a run with update_contact_agent or update_email_deliverability_agent, and read its files with get_contact_agent_file or get_email_deliverability_agent_file.

Error Handling

  • If a tool is absent, do not invent it. Call privsource_whoami and explain the missing API access.
  • Payment, NDA signing, and local file uploads stay in the dashboard. MCP returns the dashboard link and the current status.
  • If a write tool returns a payment, allowance, or credit error, stop and report the exact message.
  • Use the ids returned by PrivSource tools verbatim. Do not synthesize ids or pass numeric database ids.

Safe retries

Use an operation_key when the tool schema offers it for a message or paid operation.
Choose one key for each intended operation. Keep the same key and arguments after a lost response.
A completed repeat returns the original result. Use read tools to check current build progress.
The same key with different arguments fails. A different key starts a separate operation.
Calls without a key can repeat delivery or charges.
If an operation is in progress, retry later with the same key.
If an error persists or the outcome is uncertain, give support the operation reference.
Do not use a new key to bypass an uncertain attempt. Provider delivery is not guaranteed exactly once.

Stuck on a step?

Tell us what you need help with.

Related guides