Attio MCP and AI: Connecting Claude or ChatGPT to Your CRM

Attio's Four AI Building Blocks

Before going any further, it is necessary to distinguish four functions that are often mixed up.

The MCP server is useful when you are already working in Claude or ChatGPT. Ask Attio offers a similar experience directly within the CRM. Workflows execute processes without waiting for a manual request. Finally, Call Intelligence provides the content of sales conversations that these assistants can then leverage.

These building blocks are complementary, but their value depends on the data model, permissions, and usage rules.

Find a broader overview of the product in our Attio features guide.

What is the MCP protocol?

The Model Context Protocol, or MCP, is an open standard that allows an artificial intelligence application to access data and use external tools.

Without MCP, every connection between an assistant and software requires a specific integration. With MCP, a server exposes a list of actions in a common format: search for a record, read a note, create a task, or modify an attribute.

The assistant then chooses the appropriate action for the user's request.

For example, when you write:

Which deals over €20,000 have not been updated for two weeks?

Claude or ChatGPT can query Attio, retrieve the relevant records, and present the result in a readable response.

The MCP does not replace the API, which remains better suited for deterministic synchronizations and integrations.

What does Attio's official MCP server allow?

Attio offers an official MCP server hosted by the publisher. The connection is based on OAuth: the user logs in with their Attio account and authorizes access, without copying an API key into a configuration file.

The server can be used with Claude, ChatGPT, Cursor, and other clients compatible with remote MCP servers. In particular, it allows you to:

• search, create, and update records;

• view and manage list or pipeline entries;

• read, create, and modify notes;

• create and update tasks;

• search accessible emails and meetings;

• retrieve call recordings and transcriptions;

• read or create certain comments.

The exact list of tools may evolve. Therefore, you should check the official documentation during deployment.

The MCP does not grant unlimited access to the CRM: actions depend on the account rights, the exposed tools, and the client used.

Our Attio integrations guide details other ways to connect the CRM with Make, n8n, Zapier, or the API.

Five concrete use cases

Preparing a meeting

Before a call, you can ask:

Prepare a brief for me on Acme: contacts, last exchange, current deal, objections, and next steps.

The assistant gathers the accessible files, notes, and activities. An incomplete history will naturally produce an incomplete brief.

Updating a deal after a meeting

After an exchange, you can dictate:

Add a note on Acme with this report, move the deal to the Proposal stage, and create a task to follow up on Friday.

The assistant can prepare or execute these actions based on the confirmations requested by the MCP client.

Start by testing writing on dummy records and check the results.

Preparing a pipeline review

A sales manager can ask:

List the deals without a next task, then sort them by amount and age.

The assistant facilitates ad-hoc analysis but does not replace a reliable dashboard or properly filled-in data.

Organizing a handoff

During a handoff, the assistant can compile the available contacts, notes, opportunities, tasks, and calls, and then create a summary note to review.

Leveraging a call transcription

If Call Intelligence recorded the meeting, you can ask:

Summarize the last call with Acme, identify objections, and create follow-up tasks.

Call Intelligence captures the conversation, Attio links it to the records, and the MCP then allows you to leverage it from the assistant.

Call Intelligence: transforming meetings into CRM data

Call Intelligence allows you to record, transcribe, and store meetings in Attio. The recorder can join Google Meet, Zoom, or Microsoft Teams calls, depending on the workspace conditions and settings.

Users can then view the video, transcription, and AI-generated insights. Insight templates allow for extracting important information based on context:

• next steps;

• objections;

• budget discussions;

• technical requirements;

• risks;

• product feedback.

Attio links meetings to the relevant people and companies. Call Intelligence is currently reserved for Pro and Enterprise plans and must therefore be factored into the project cost.

Check out our article on Attio pricing to compare plans.

Ask Attio, AI attributes, and agents

The MCP server is only one part of Attio's AI strategy.

Ask Attio

Ask Attio allows you to query the CRM directly from its interface. It can search for information, summarize records, create tasks, and act on accessible data.

The user stays within Attio, whereas MCP is better suited for teams already working in Claude or ChatGPT.

AI attributes

AI attributes allow you to generate or complete values on records and lists.

For example, they can:

• summarize a company;

• classify a prospect;

• clean up a value;

• extract information;

• produce structured qualification.

The result thus becomes visible, filterable, and usable in automations. These attributes are available on all plans, with credit volumes varying by subscription.

Agents in Workflows

Attio Workflows start from a trigger: record creation, attribute modification, scheduled deadline, or external webhook.

They can then execute several steps, conditions, and actions. AI agents allow for adding analysis or transformation to this process.

For example:

1. a new company enters a list;

2. an agent researches its activity;

3. the result is structured into several attributes;

4. the record is assigned to the correct person;

5. a task is created if the company matches the criteria.

A Workflow is preferable to MCP when an action needs to be triggered automatically and repeatably, without waiting for a user conversation.

What the MCP server does not replace

The MCP makes it easier to access Attio, but it does not fix a poorly designed CRM.

It does not replace:

• a coherent data model;

• deduplication rules;

• appropriate permissions;

• properly named attributes;

• an event-driven Workflow;

• a deterministic integration;

• human validation;

• a data retention policy;

• user training.

An instruction in natural language can remain ambiguous. For critical actions—deletion, owner change, amount, or external sending—keep human confirmation.

Security, permissions, and GDPR

Connecting an AI to a CRM gives access to business and personal information. Implementation must therefore be treated as a governance matter, not just a technical setting.

Using the official server

The official server with OAuth avoids sharing a static API key. However, accounts must be protected, strong authentication enabled, and authorized applications monitored.

Checking Attio permissions

The MCP acts according to the permissions of the connected account. Therefore, access rights must be reviewed before widespread usage.

Detailed controls vary by plan. Ensure that each member only sees the information necessary for their role. Permissions on objects and lists are more granular on Pro and Enterprise.

Starting with read-only

Start with read-only use cases. Then introduce write actions with:

• dummy records;

• confirmation before execution;

• verification in Attio;

• a procedure to report and correct errors.

Managing recordings

Call recordings deserve special attention. Attio currently indicates that all recordings are visible to all members of the workspace and can also be shared externally. This visibility is not individually configurable.

Before enabling Call Intelligence, define:

• which calls can be recorded;

• meetings that must remain excluded;

• retention period;

• deletion conditions;

• who can access the workspace.

Specifically, avoid recording HR, legal, or financial exchanges in the same space that should not be accessible to the entire team.

Informing participants

In France and the European Union, the individuals concerned must be informed of the recording and processing of their data. The CNIL emphasizes the need to explain the purpose of the operation, the recipients, and the applicable rights.

The legal basis depends on the context; consent is not always the only option. Have this framework validated by your GDPR officer or legal counsel.

How to start without taking risks

The connection is quick, but deploying to a team requires more preparation.

1. Clean up the testing scope

Choose a pipeline or a few non-sensitive records. Check for duplicates, attributes, and permissions.

2. Connect Attio

In Claude, ChatGPT, or another compatible client, add the Attio connector and follow the OAuth authentication.

3. Test read-only first

Start by asking for:

• the list of open deals;

• a summary of an account;

• upcoming tasks;

• the last available call.

Verify each response in Attio.

4. Introduce write actions

Next, test creating a note and a task, then a simple modification on a dummy record.

5. Define internal rules

Document authorized uses, actions requiring validation, excluded data, and the correction procedure.

6. Expand progressively

Once usage is stable, transform repetitive scenarios into Workflows or robust integrations.

Need help building a truly AI-augmented CRM?

HyperOps helps B2B companies design, automate, and govern their Attio CRM.

We structure the data model, connect the right tools, deploy Workflows, and oversee MCP usage. The goal is not to add AI everywhere, but to reduce data entry and make data more reliable.

Discover our support on the Attio Agency page, our expertise in automation and AI, or tell us about your project.

Frequently Asked Questions on Attio MCP and AI