# Attio MCP > Attio's MCP server connects your CRM to the AI tools your team already has open. Sign in once from ChatGPT, Claude, Notion, or anything else that speaks MCP, then search records, log notes, create tasks, and update deals without switching tabs. Your AI tools can read from Attio and write back to it, over Universal Context — the deepest context layer in agentic CRM — with access scoped to each person and actions verified before they run. ## What is Attio's MCP server? MCP (Model Context Protocol) is an open standard for connecting AI tools to the systems they need. Attio's MCP server makes your workspace available to any MCP client, so the tool you already have open can read from Attio and write back to it. It combines three things: - **Speaks your language.** No views to build, no filters to set. Get insights and take action in conversation. - **Full workspace access.** Search, create, and update across your CRM records and activities. - **Secure by design.** Authenticate via OAuth with user-scoped permissions. The result: your CRM stops being a place you always have to go and becomes something your AI tools can always reach, in every tool your team already works in. ## Developer Quotes - "The Attio MCP gave us instant access to Attio's full data model without custom API work." — Siavash Ghorbani, Stilla ## Scale - **471K+** — average MCP calls made from Attio workspaces every day - **40%** — of Enterprise workspaces use Attio MCP ## Core Capabilities ### Search & Retrieve - Ask questions about any record, company, or contact in plain language - Run semantic search across notes, emails, and call recordings - Pull a full account history before a meeting without opening Attio ### Create & Update - Update deal stages, values, and next steps as conversations happen - Log notes against any record the moment a call ends - Create and update tasks, assign owners, and set due dates - Create and update records across any object in your workspace - Build a target list and add records to it as you research. ### Meeting & Call Intelligence - Search meetings and call recordings by metadata or meaning - Retrieve a specific call recording or transcript - Find the conversation where a topic came up, then read exactly what was said - Build lists of quotes and themes across many recorded conversations ### Research & Analysis - Surface patterns across accounts rather than one record at a time - Assemble reports on your interactions with any company - Extract product feedback and objections from calls at scale ## Universal Context™ ### The context layer behind every connection Lots of CRMs can open a connection. What comes back is the difference. - **The deepest context in agentic CRM** — calls, emails, notes, meetings, and product usage, all within reach of the tool you connect - **Answers from your records** — every response traces back to real data in your workspace - **Built for AI to use** — structured so tools can find what they need and act on it - **Signals across every account** — see how things trend over time, beyond a single moment - **Your whole stack, connected** — Granola notes, Slack threads, and whatever you adopt next ## Use Cases ### Sales - Update the deal from the tool where you took your notes, the moment the call ends - Walk into every meeting briefed, with prior calls, emails, and notes pulled in seconds - Create follow-up tasks and reminders for next steps on key deals - Research an account and write the findings back to the record ### Customer Success - Pull account health and history before an onboarding call or QBR - Find every conversation where a customer raised a blocker or a request - Log notes and next steps against the account without leaving your tool - Build a handoff brief from prior conversations and product usage ### Marketing - Build lists of customer quotes and feedback themes from recorded calls - Give teams who do not live in the CRM a way to reach customer data - Pull aggregate stats on your customer base by location, industry, or segment - Find customers with a particular use case for case studies, speakers, or events - Surface the language customers actually use to sharpen positioning and messaging ### Revenue Operations - Query pipeline and account data in plain language, without building a view - Analyze deals against your own filter criteria to consolidate trends and forecasting - Assemble a report of every interaction with a given account - Track engagement across customers and prospects ## Which tools can connect ChatGPT, Claude, Notion, and anything else that speaks MCP. Some clients, like Claude and ChatGPT, offer one-click installation from their built-in connector directories. Any other MCP client can connect using the custom connector URL, https://mcp.attio.com/mcp. Because MCP is an open standard, tools you adopt later will work without new setup. ## Getting connected 1. **Find Attio.** Select Attio from the connector directory in the tool you want to work from, or add https://mcp.attio.com/mcp as a custom connector. 2. **Sign in.** Authenticate with OAuth using the Attio account you already have. No API keys required. 3. **Start working.** Access follows what you can already see and do in Attio. No setup, no code, and no waiting on anyone else. ## Security & Permissions ### Yours to control - **OAuth by default** — log in with your Attio account, no API keys to manage - **Per person, not per workspace** — you have the same permissions as your Attio user account - **Scoped to your workspace** — operations reach workspace data only - **Reads flow, writes confirm** — read operations are auto-approved; creating or updating data requests your confirmation first - **Revocable anytime** — end any session from your Attio account settings ## Common Questions **How is this different from Ask Attio?** Ask Attio is Attio's own assistant, working inside your workspace. The MCP server does the reverse: it makes your Attio workspace available to AI tools outside Attio, so you can work from the tool you already have open. They are separate parts of Attio's AI story and can be used together. **How is this different from the API?** The API is for developers building integrations and custom software. The MCP server needs no code: you connect it from any AI tool in a few clicks, and the tool discovers what it can do on its own. **Does it only read data, or can it make changes?** Both. The server supports read and write, so a connected tool can create records, update deals, log notes, and create tasks, not just retrieve information. **How do permissions work?** Any active workspace member can connect through OAuth. Access works per individual user rather than per workspace, and mirrors that person's existing permissions in Attio. **Which tools can I connect from?** Any MCP client, including ChatGPT, Claude, and Notion. Because MCP is an open standard, new tools work without new setup. ## What the server exposes - **Records and objects** — search records, list records, get records by ID, create, update, upsert, and merge records, list attribute definitions - **Lists** — list all lists, list entries in a list, add a record to a list, update lists and list entries - **Notes** — create, search by metadata, semantic search, get full note body, update - **Tasks** — list, create, and update tasks with deadlines, assignees, and linked records - **Meetings and calls** — search meetings, search call recordings by metadata or meaning, retrieve a recording with its full transcript - **Emails** — search by metadata or meaning, retrieve full email content - **Comments** — create, list, reply to, and delete comments on records and list entries - **Reporting** — run aggregate reports on records or list entries, with filters and grouping - **Workspace** — list members, list teams, confirm current user identity Full reference: https://docs.attio.com/mcp/overview ## Getting Started - Start for free - MCP product page - MCP documentation - Help center ## Related pages - Ask Attio - AI-powered features - Developer platform - Automations & workflows - Data structure & syncing - Complete platform overview