Pipedrive MCP server

What is an MCP server for Pipedrive?

Pipedrive is the CRM of choice for many small and mid-sized sales teams: lean, visual, focused on the pipeline. An MCP server for Pipedrive starts exactly there. It translates Pipedrive's functions into standardised tools that an AI agent can call through the Model Context Protocol, whether Claude, GPT, Gemini or Copilot. Instead of writing a separate script for each AI integration, all MCP-capable models talk to the same interface.

The foundation is solid: Pipedrive provides a documented REST API (currently in transition from v1 to the newer v2) with authentication via OAuth 2.0 or API token. Your own MCP server sits in between as a translation layer: it takes the agent's request, calls the matching API endpoints and returns the result in structured form.

Is there an official Pipedrive MCP server?

Yes. On 30 June 2026, Pipedrive introduced a native MCP server (Pipedrive press release). According to Pipedrive, it is available on all plans, is set up through an OAuth connection and works with ChatGPT, Claude and other AI assistants that support their own MCP connections. With it, the agent can search deals, people, organisations and leads, analyse pipelines, create and update records, convert leads into deals and manage activities.

For many teams, the official server is the obvious starting point. We build your own MCP server on the basis of the official REST API when you need more control: your own tools tailored to your pipeline, firmly limited rights per agent, approval steps before actions that change data, your own audit log, or operation with a model of your choice on your own infrastructure. We clarify which route fits in the AI Check.

Use cases: what can a Pipedrive MCP server do in practice?

1. Creating deals from a conversation After a first call you dictate the key details, and the agent creates the deal in the right pipeline and stage, with the linked person, organisation and estimated value. No switching to the form, no forgotten leads.
2. Logging activities without gaps Calls, emails and meetings often only reach the CRM days later, if at all. The agent records them as an activity directly on the matching deal and plans the next follow-up action along with it.
3. Pipeline hygiene & stale deals “Which deals in the negotiation stage haven't moved in two weeks?” The agent pulls the list, proposes reactivation steps and creates the reminders, before opportunities fizzle out.
4. Forecast & reporting on request “What is my weighted forecast for this month?” or “Which organisation brings the most open volume?” The agent reads the figures from Pipedrive without anyone clicking filters and views together.
5. Lead qualification & enrichment Incoming leads are scored against your criteria, matched with existing people and organisations and given notes. The agent reports duplicates instead of blindly creating them.
6. Keeping master data clean The agent checks where people or organisations are missing fields such as email, phone number or VAT ID and proposes additions, including by reconciling with adjacent systems such as the newsletter tool or accounting.

How the integration works

With your own MCP server, the architecture overview is deliberately lean:

AI agent ↔ MCP server (your EU host) ↔ Pipedrive REST API (OAuth 2.0, REST/JSON)

In concrete terms, we proceed as follows:

• OAuth connection: You authorise the MCP server once for your Pipedrive account. The tokens are stored encrypted on the EU host, including a rotation and revocation strategy.
• Tool mapping: We define MCP tools such as create_deal, list_deals, log_activity, get_person and update_deal_stage, each with a JSON schema, validated inputs and understandable error messages.
• Custom fields: Your own fields are mapped as well, so the agent does not fail on a generic standard but knows your actual pipeline logic.
• Read vs. write separation: By default, the agent reads and drafts. Actions that change data (stage change, closing a deal, deleting) go through an approval step that you configure.
• Logging: Every API action is logged with timestamp, triggering user and change details, so it is always traceable what the agent did.

GDPR, hosting & security

A CRM holds the most sensitive data a sales team has: contacts, terms, conversation notes. An AI integration must not weaken this confidentiality. Pipedrive itself offers EU data centres, which is a good starting point. With the official MCP server, Pipedrive runs the connection layer itself; where it runs and which terms apply, we check before use against Pipedrive's contract documents. With your own MCP server, we make sure of the following:

• EU hosting of the MCP server (for example Hetzner or the Azure region Germany West Central in Frankfurt), without US edge nodes in the data path.
• Encrypted OAuth tokens with defined rotation and the ability to revoke access at any time.
• AI model with EU data residency and a data processing agreement (DPA), for example Claude via AWS Bedrock (EU region Frankfurt), Google Vertex AI (EU), OpenAI Enterprise EU or Azure OpenAI in an EU region.
• Data minimisation: The agent sees only the fields and pipelines it needs for the respective task, with no blanket access to the whole account.
• Role and visibility boundaries, so that an agent for the sales team does not accidentally reach data from another area.

What does it cost?

Pipedrive is among the faster MCP projects because the REST API is well documented and the data model stays manageable. In concrete terms:

• Start small: usually with deal creation or activity logging as the starting point. We set the timeline after the AI Check.
• Clear scope per use case instead of open-ended hourly billing, then expansion in clearly delimited steps.

We name a reliable figure after the initial consultation, depending on pipeline complexity, the tools you want and whether adjacent systems are connected too.

Frequently asked questions

Is there an official MCP server from Pipedrive?

Yes. On 30 June 2026, Pipedrive introduced a native MCP server. According to Pipedrive, it is available on all plans, is connected via OAuth and can search deals, people, organisations and leads, analyse pipelines and create and update records. We also build your own MCP server on the official REST API if you need your own tools, firmly limited rights, approval steps or your own audit log.

Which Pipedrive objects can the AI agent reach?

Deals, people, organisations, activities, leads and pipelines, including your custom fields. Depending on permissions, the agent can read, create or update these.

Does the AI agent write to my deals without being asked?

Only if you want it to. By default we separate read and write permissions. Actions that change data, such as a stage change or closing a deal, go through an approval step so that you keep control.

Is a Pipedrive MCP integration GDPR-compliant?

It can be if all links in the chain are right: MCP server in the EU, encrypted tokens, an AI model with EU data residency and a data processing agreement. Pipedrive offers EU data centres for this. We design the integration with this requirement from the start.

How long does setup take?

That depends on the route: the official MCP server is connected via OAuth, while your own server depends on scope and approval logic. We set the timeline after the AI Check and start with a clearly defined use case.

Which AI models work with the Pipedrive MCP server?

All MCP-compatible models: Anthropic Claude, OpenAI GPT, Google Gemini and Microsoft Copilot. A later model change is possible without a new integration.