An Attio MCP that knows your workspace has its own schema
Connect one Attio workspace API key and 76 nodes land on the canvas: query and upsert records on any object, move records through lists with their own stage attributes, read and replace attribute value history, write notes and tasks, comment on records, sync meetings and call recordings with transcripts, manage files and folders, and register webhooks. Because Attio has no fixed schema, the nodes discover your objects, attributes and lists at run time rather than assuming a shape.
What the Attio MCP connector does
The Attio connector for Mattering Automations is 76 nodes on Attio's v2 REST API: 38 read nodes and 38 action nodes. It covers objects and their attributes, select options and statuses, records with query, get, create, upsert, update, overwrite, merge and delete, attribute value history on both records and list entries, lists and list entries with their own stage and owner attributes, saved views on both, fuzzy record search, read only SQL, workspace members, notes, tasks, comment threads, email metadata, meetings, call recordings with transcripts, files and folders, and webhook registration. One workspace API key authenticates all of it, sent as a bearer token.
Every node speaks MCP, the open protocol for connecting AI assistants to tools, so Claude, ChatGPT, and any MCP capable assistant can use them, alongside Matter, the AI assistant built into the platform. An agent can find a company with Attio: Search Records, write a research brief onto it with Attio: Create Note, put it into a pipeline with Attio: Upsert List Entry, and open a follow up with Attio: Create Task, all in one pass and all recorded in Attio rather than in a chat log.
One structural fact about Attio shapes the whole connector, and it is the reason this is not a generic CRM integration. Attio has no fixed schema: objects, attributes, lists and entries are all workspace defined. There is a people object because Attio ships one, and your workspace can add a projects object, rename it, and hang new attributes off it whenever it likes. So the interesting identifiers cannot be baked in, they have to be discovered. The nodes are built around four dependent chains that do exactly that: object to attribute, object to attribute to select option or status, list to entry, and object to record. Choose an object and the attribute picker offers only that object's attributes; choose an attribute and the option and status pickers narrow again. Record pickers read from Attio: Search Records rather than from the record lister for a specific reason worth knowing: a record's values arrive as a map keyed by attribute slug, which no path expression can fan out over, so the lister cannot supply a display name and search is the one place Attio hands back the name it renders itself. The cost of that choice is stated on every field that uses it, because search is documented as beta and eventually consistent, so a record created seconds ago may not be findable yet.
The safety modeling is stated rather than implied. 11 of the 76 nodes are flagged destructive, and Attio's own archiving guide is unambiguous that deleted data is eventually removed entirely with no restoration: no trash, no restore window, no undo. Attio: Delete Record takes the record, every attribute value and its history, and the record's entries in every list. Attio: Merge Records destroys the secondary record and resolves every conflict in the primary's favour with no unmerge. Attio: Replace Record Attribute History and Attio: Replace Entry Attribute History destroy an attribute's entire value history, including values not in your request, which on a list entry is the deal's stage history. Attio: Delete Comment has the least obvious blast radius: deleting the FIRST comment in a thread deletes every other comment in that thread, including replies by other people. Attio: Delete File deletes every descendant when the target is a folder. Nothing in the attributes family is destructive, because Attio ships no delete for attributes, options or statuses at all: they are archived instead, which is reversible, and the update nodes carry that flag.
One connection, three surfaces
- Visual automations. Drag Attio nodes onto the canvas and wire them to triggers, conditions, and other apps.
- AI agents, yours or ours. These nodes speak MCP, the open protocol, so Claude, ChatGPT, and any MCP-capable assistant can use them, and Matter, the agent built into the platform, uses them natively.
- The whole platform. Combine with native tasks, projects, and other connectors, no middleware in between.
Every Attio node, ready to drop on the canvas
76 nodes, generated from the connector itself so this list always matches what ships. Expand a category to browse.
Read and query nodes 38
Attio: Get Attribute
Get one attribute's full definition: type, whether it is writable, required, unique or multiselect, its default value, its type-specific config, and for a record-reference attribute the reverse relationship it is entangled with.
Attio: Get Call Recording
Get one call recording with its full transcript: timed segments with speaker attribution at data.data.transcript.segments[], plus a flat readable version at data.data.transcript.raw_transcript, and the video URL if there is one. This is the node to feed a summariser. Documented as beta.
Attio: Get Comment
Get one comment by id, with its author, text, thread, the record or entry it sits on, and whether it has been resolved.
Attio: Get File
Get one file or folder entry by id, with its name, content type and size, its storage provider, and the record it hangs off. Returns the metadata, not the bytes. Documented as beta.
Attio: Get List
Get one list by slug or id, with its parent object and the workspace and per-member access levels. Read the parent object before adding entries: a list only accepts records of that object.
Attio: Get List Entry
Get one list entry by id, with its parent record and every list-specific attribute value. Read data.data.parent_record_id to jump from the entry back to the record it represents.
Attio: Get Meeting
Get one meeting by id with its title, description, start and end, participants and their RSVP status, and linked records. Documented as beta.
Attio: Get Note
Get one note by id, in both plaintext and markdown, with the record it belongs to, the meeting it is linked to if any, and the records and members it tags.
Attio: Get Object
Get one object by slug or id, with its api_slug, singular and plural nouns and creation time. Use it to resolve a slug to the stable object_id before saving a reference that must survive someone renaming the object in the UI.
Attio: Get Record
Get one record by id, with every attribute's current value and the record's web_url for linking a human straight to it. As with the lister, data.values is keyed by attribute slug, so a company's name reads as data.data.values.name[0].value.
Attio: Get Task
Get one task by id, with its content, deadline, completion state, assignees and the records it links to.
Attio: Get Thread
Get one thread with every comment in it, each with its author, text, creation time and whether the thread has been resolved.
Attio: Get Webhook
Get one webhook with its target URL, subscriptions, filters and status. The signing secret is not echoed.
Attio: Get Workspace Member
Get one workspace member by id. Useful for turning the actor id that appears all over Attio's responses (created_by_actor.id) into a name and an email address a notification can address.
Attio: Identify Token
Introspect the access token: whether it is active, the scopes it carries, and the workspace it belongs to by id, name and slug. This is the cheapest authenticated read Attio offers and the only endpoint that needs no scopes at all, which makes it the right connection test. Note that a revoked token can answer HTTP 200 with active false rather than 401, so check data.active, not just the status code.
Attio: List All Lists
List every list the token can see, in sidebar order, with each one's slug, name, parent object and access settings. A list in Attio is a pipeline or a working set: records are added to it as ENTRIES, and each entry carries its own attributes such as stage or owner. This is what every list picker reads.
Attio: List Attributes
List the attributes defined on one object or one list, in the order the UI sorts them, with each attribute's slug, title, type and whether it is required, unique, multiselect or a system attribute. Because Attio's schema is workspace-defined, this is how you discover the keys a record's values object will accept, and it is what every attribute picker reads.
Attio: List Call Recordings
List the call recordings attached to a meeting, with each one's processing status and a web URL into Attio. The transcript itself is not in the list payload; fetch one recording to get it. Documented as beta. Cursor-paginated.
Attio: List Emails
List email METADATA from the workspace's connected mailboxes: who, when, which direction, what subject line, and which records the message is linked to. Message content is never returned. Three things to know before wiring it up. It is documented as ALPHA and is enabled per workspace and per app on request to Attio support, so it can 403 on a token that works everywhere else. It refuses to list everything: supply at least one of linked records, participants or domain. And filters that identify your own workspace are ignored, so filtering by a colleague's address returns nothing.
Attio: List Entries
Query the entries in one list with a filter and a sort order. An entry is a record's membership of that list plus the list-specific attributes attached to it, so this is where pipeline stage, owner and deal-specific fields live. As with records, data.data[].entry_values is keyed by attribute slug rather than being a list. POST because the filter goes in the body; nothing is written.
Attio: List Entry Attribute Values
Read one attribute's values on one list entry, optionally including the full history oldest first. This is how you reconstruct a deal's stage history: each value carries active_from and active_until, so the gaps between them are the time the entry spent in each stage.
Attio: List Files
List the files and folders attached to ONE record: both files stored in Attio itself and folders connected from Dropbox, Box, Google Drive or OneDrive. The object and record id are required, so there is no way to list a workspace's files as a whole. Documented as beta. Cursor-paginated.
Attio: List List Views
List the saved views on a list, ordered by view id. Feed a view id to Attio: List Entries to reproduce exactly the filtered board a person sees in the UI. Cursor-paginated.
Attio: List Meetings
List meetings with their participants, times and linked records, filtered by date window, participant or linked record. Note that participants and linked_record_id are combined with OR, not AND, so supplying both widens the result rather than narrowing it. Documented as beta. Cursor-paginated.
Attio: List Notes
List notes across the whole workspace, or narrow to one record by supplying both the parent object and the parent record id. Each note comes back in both plaintext and markdown, with the records and members it tags, which makes this a good feed for summarisation.
Attio: List Object Views
List the saved views on an object, ordered by view id. A view carries a filter configuration someone built in the UI, so feeding a view id to Attio: List Records reproduces exactly what a person sees on screen without rebuilding the filter in JSON. Cursor-paginated.
Attio: List Objects
List every object in the workspace, both the ones Attio ships (people, companies, deals, users, workspaces) and any the workspace has defined itself, with each one's api_slug and singular and plural nouns. Attio has no fixed schema, so this is the node everything else leans on: it is what an object picker reads, and running it is how you find out what a workspace actually models.
Attio: List Record Attribute Values
Read one attribute's values on one record, optionally including every historic value rather than just the current one. Attio keeps value history with active_from and active_until timestamps, so this is how you answer questions like when a deal changed stage or what a job title used to be. Historic values are returned oldest first, and are not available for communication-intelligence or enriched attributes, which return 400 instead.
Attio: List Record Entries
List every list entry, across every list, whose parent is this record. This is how you answer which pipelines a company is currently in, and it returns each list's id and slug alongside the entry id so a follow-up node can read or update the entry.
Attio: List Records
Query records of one object with a filter and a sort order. This is the workhorse read: strongly consistent, unlike search, so it is what to use when a workflow has just written something and needs to see it. Note that data.data[].values is an object keyed by attribute slug rather than a list, so a person's name reads as data.data[0].values.name[0].full_name. It is a POST because the filter goes in the body; nothing is written.
Attio: List Select Options
List the options a select attribute offers, with each option's title, id and archived flag. Run this before writing to a select attribute: Attio matches on the option title, so a typo silently fails rather than creating a new option.
Attio: List Statuses
List the statuses a status attribute offers, with each one's title, id, celebration flag and target time in status. This is the pipeline vocabulary for a list or object, so run it before moving anything between stages.
Attio: List Tasks
List tasks, filtered by assignee, completion state, or the record they are linked to, and sorted by creation or completion time. This is the read behind an overdue-task digest or a per-owner workload report.
Attio: List Threads
List comment threads, optionally narrowed to one record (pass the object and record id) or one list entry (pass the list and entry id). Each thread arrives with its comments nested inside, so a single call gives you the whole conversation at data.data[].comments[]. This is also the only lister that surfaces comment ids.
Attio: List Webhooks
List the workspace's webhooks with their target URLs, event subscriptions and status. The signing secret is returned only when a webhook is created, never here, so this is safe to run in a shared workflow.
Attio: List Workspace Members
List everyone in the workspace with their name, email address, avatar and access level. Workspace members are the only actors that can own a task or author a comment, so this is what every assignee and author picker reads. The API is read-only here: members are invited and removed in the Attio UI.
Attio: Query SQL
Run a read-only SQL query across records and lists and get back flat rows, which is the one node here whose output drops straight into a chart or a spreadsheet without walking Attio's value envelopes. ENTERPRISE ONLY: Attio gates this endpoint to Enterprise workspaces and any other plan gets an error, and it is documented as beta.
Attio: Search Records
Fuzzy-search records across one or more objects the way the Attio search bar does, matching names, domains, email addresses, phone numbers and social handles on people and companies, and labels on everything else. It returns a record_text display name for each hit, which is why every record picker in this connector reads from here rather than from the record lister. Two caveats: it is documented as beta, and results are EVENTUALLY CONSISTENT, so use Attio: List Records when you need to see a write you just made.
Action nodes 38
Attio: Add Record To List
Add an existing record to a list as a new entry, setting the list's own attributes at the same time. This ALWAYS creates, and Attio explicitly allows several entries for the same parent record, so calling it twice gives you two entries. Use Attio: Upsert List Entry when a workflow might run twice.
Attio: Create Attribute
Add an attribute to an object or a list. The type decides what config the attribute accepts and cannot be changed afterwards. For a record-reference attribute you can pass a relationship object to create a bidirectional pair, which creates a second, entangled attribute on the other object. Attributes are archived rather than deleted, so this is reversible in the sense that it can be hidden, not removed.
Attio: Create Call Recording
Attach a recording and transcript to a meeting, which is how an external notetaker gets its output into Attio. Always send the transcript: it is optional only for backwards compatibility, and a recording without one has no summaries and none of the transcript-derived features. The video is optional and Attio downloads it from your URL asynchronously. RATE LIMITED TO ONE REQUEST PER SECOND, well below the general write limit, so throttle a backfill. Documented as beta.
Attio: Create Comment
Post a comment. Supply EXACTLY ONE of thread_id (reply to an existing thread), record (start a thread on a record) or entry (start a thread on a list entry). Putting a workspace member's email address in the content NOTIFIES them, which is the mechanism for pinging an owner from a workflow; any other email address just becomes a mailto link. The author must be a workspace member, because Attio does not yet support other actor types here.
Attio: Create Folder
Create a folder on a record, or connect a file or folder that already lives in Dropbox, Box, Google Drive or OneDrive. For a native folder set file_type to folder and give it a name. To connect something external, set file_type to connected-folder or connected-file and supply the storage provider and that provider's own file id. Documented as beta. Note this body is NOT wrapped in a data envelope, unlike most Attio writes.
Attio: Create List
Create a list and set who can see it. Attio requires that SOMEONE has full access: either workspace_access is full-access, or at least one entry in workspace_member_access has level full-access, otherwise the call is rejected. Per-member and advanced workspace access are plan-gated, so a workspace on a lower plan can get a 403 here. Add attributes afterwards with Attio: Create Attribute (target lists) and records with Attio: Create List Entry.
Attio: Create Meeting
Record a meeting in Attio, which is how an external calendar or notetaker syncs its events in. Participants given by email address AUTOMATICALLY CREATE person and company records if they do not exist, so this node can grow the CRM as a side effect; participants given by name only do not. Attio has no built-in de-duplication here, so a re-run creates a second meeting. Documented as beta.
Attio: Create Note
Write a note onto a record. This is the natural landing place for an AI-generated call summary or research brief: it accepts markdown, so headings and lists survive, and it can be backdated with created_at when migrating history in.
Attio: Create Object
Create a custom object, which is Attio's equivalent of adding a new record type to the CRM. A new object starts with only its system attributes; add your own with Attio: Create Attribute. Attio ships no endpoint that deletes an object, so treat this as close to permanent even though it is not marked destructive.
Attio: Create Record
Create a record. This ALWAYS creates: it throws on a conflict with a unique attribute rather than updating, so a workflow that might see the same person twice wants Attio: Upsert Record instead. The values object is keyed by attribute slug or id, and which keys are valid depends entirely on how the workspace has configured that object.
Attio: Create Select Option
Add an option to a select attribute. Options are archived rather than deleted, so this only ever grows the vocabulary.
Attio: Create Status
Add a status to a status attribute, which is how you add a stage to a pipeline. Statuses are archived rather than deleted.
Attio: Create Task
Create a task, assign it and link it to records. The linked records array is unusually forgiving: it accepts a record id, a company domain, a person's email address, or a unique matching attribute, so a workflow can link a task to a person it only knows by email without looking the record up first. Content is PLAINTEXT ONLY at present, with no rich text, links or at-mentions.
Attio: Create Webhook
Register a webhook so Attio pushes record, list-entry, note, task, comment and attribute events to a URL. THE RESPONSE IS THE ONLY TIME THE SIGNING SECRET IS RETURNED (data.data.secret): capture it in the same run or you have to recreate the webhook. Each combination of target URL, event type and filter must be unique in the workspace; a duplicate is rejected with 409.
Attio: Delete Call Recording
DESTRUCTIVE AND IRREVERSIBLE. Delete a call recording and everything derived from it: the video, the transcript, and the summaries Attio generated from that transcript. The meeting itself survives. Unless you still hold the source recording elsewhere, this is unrecoverable. Answers HTTP 204 with no body. Documented as beta.
Attio: Delete Comment
DESTRUCTIVE AND IRREVERSIBLE. Delete a comment. Blast radius worth reading twice: if the comment is the FIRST one in its thread, Attio deletes EVERY OTHER COMMENT in that thread too, including replies written by other people. Deleting a reply only removes that reply. There is no undo. Returns an empty object.
Attio: Delete File
DESTRUCTIVE AND IRREVERSIBLE. Delete a file or folder entry. DELETING A FOLDER DELETES EVERY DESCENDANT, files and subfolders alike, in one call and with no confirmation. For a file stored in Attio the bytes go with it; for a CONNECTED entry only Attio's link is removed and the file stays in Dropbox, Box, Drive or OneDrive. Check the file_type and storage_provider before running this. Documented as beta. Returns an empty object.
Attio: Delete List Entry
DESTRUCTIVE AND IRREVERSIBLE. Remove an entry from a list, taking with it every list-specific attribute value on that entry and their history: the stage it reached, when it got there, who owned it. The underlying RECORD survives untouched, which is what makes this the right way to take something out of a pipeline rather than deleting the record. There is no undo, and re-adding the record creates a fresh entry with empty values. Returns an empty object.
Attio: Delete Note
DESTRUCTIVE AND IRREVERSIBLE. Delete a note and its content. Attio has no trash for notes and no restore, so a call summary deleted here is gone. The record it was attached to is unaffected. Returns an empty object.
Attio: Delete Record
DESTRUCTIVE AND IRREVERSIBLE. Delete a record and everything that hangs off it: every attribute value and its history, and the record's entries in every list it appears on. Attio has no trash and no restore window, and its own guidance is that deleted data is eventually removed from its servers entirely. Notes, tasks and comments that reference the record lose that reference. If what you want is to take a record out of a pipeline, delete its LIST ENTRY instead. Returns an empty object.
Attio: Delete Task
DESTRUCTIVE AND IRREVERSIBLE. Delete a task. There is no undo and no trash. If what you want is to mark work finished, set is_completed true with Attio: Update Task instead, which keeps the record of it. Returns an empty object.
Attio: Delete Webhook
DESTRUCTIVE AND IRREVERSIBLE. Delete a webhook and every subscription on it. Delivery stops immediately and the signing secret is destroyed with it, so recreating the webhook means distributing a NEW secret to whatever verifies the signatures. Events that happen while it is gone are not replayed. Returns an empty object.
Attio: Merge Records
DESTRUCTIVE AND IRREVERSIBLE. Merge two records into one. The SECONDARY record ceases to exist: its values survive only where the primary had none, and every conflict is resolved in the primary's favour, so the secondary's version of any contested attribute is lost. There is no unmerge. Run Attio: Get Record on both first if you want a copy of what is about to be discarded.
Attio: Overwrite List Entry Values
Update a list entry, OVERWRITING multiselect attributes: for every multiselect attribute you mention the supplied list becomes the entire list, so anything omitted is removed. Attributes you do not mention are untouched. Use Attio: Update List Entry to add without removing.
Attio: Overwrite Record Values
Update a record, OVERWRITING multiselect attributes. For every multiselect attribute you mention, the supplied list becomes the entire list and anything not in it is removed, which is how you clear tags or replace a set of linked records. Attributes you do not mention are untouched. Use Attio: Update Record to add without removing.
Attio: Replace Entry Attribute History
DESTRUCTIVE AND IRREVERSIBLE. Replace a list entry attribute's ENTIRE value history. Every value it currently holds is destroyed, including values not in your request, and Attio requires replace_history true as an acknowledgement. It exists for migrating stage history in from another CRM, not for moving an entry between stages, which belongs on Attio: Update List Entry. These writes fire NO webhooks and NO workflow triggers, and history cannot be written for relationship or formula attributes.
Attio: Replace Record Attribute History
DESTRUCTIVE AND IRREVERSIBLE. Replace an attribute's ENTIRE value history on one record. Every value the attribute holds today is destroyed, including values not present in your request, and only what you send survives. Attio requires replace_history true as an explicit acknowledgement of that. It exists for migrating history in from another system, not for ordinary edits, which belong on Attio: Update Record. Note also that these writes fire NO webhooks and NO workflow triggers, so a migration does not replay automations, and that history cannot be written for relationship or formula attributes.
Attio: Update Attribute
Rename an attribute, change its constraints or default, or ARCHIVE it by setting is_archived true. Archiving is Attio's answer to deletion here: the attribute and its data are hidden rather than destroyed, and setting is_archived back to false restores them. The attribute's type cannot be changed. Fields left blank are not touched.
Attio: Update List
Rename a list or change who can see it. The parent object cannot be changed through the API at all: Attio deliberately restricts that to the UI because of the side effects. The same someone-must-have-full-access rule as creation applies, and sending workspace_member_access REPLACES the whole access list rather than adding to it.
Attio: Update List Entry
Update a list entry, APPENDING to multiselect attributes: values you send are prepended to what is already there and nothing is removed. Attributes you do not mention are untouched. This is the node for moving an entry to a new stage. Use Attio: Overwrite List Entry Values when the supplied list should become the whole list.
Attio: Update Object
Rename an object or change its api_slug. Changing the slug breaks anything that referenced the object by its old slug, which is exactly why Attio recommends storing the object_id when an integration has to survive renames. Fields left blank are not changed.
Attio: Update Record
Update a record, APPENDING to multiselect attributes. Values sent for a multiselect attribute are added in front of the ones already there; nothing is removed. Attributes you do not mention are untouched. Use Attio: Overwrite Record when you need the supplied list to be the whole list.
Attio: Update Select Option
Rename a select option or archive it. Archiving hides the option from new selections without removing it from records that already carry it, and it is reversible. Renaming changes the title everywhere at once.
Attio: Update Status
Rename a status, change its target time or celebration setting, or archive it. Archiving removes the stage from new selections without touching records that already sit in it, and is reversible.
Attio: Update Task
Change a task's deadline, completion state, links or assignees. The task's CONTENT cannot be edited through the API, which is Attio's limitation rather than this node's. Sending linked_records or assignees REPLACES that list wholesale, so read the task first if you mean to add rather than replace.
Attio: Update Webhook
Change a webhook's target URL or its subscriptions. Sending subscriptions REPLACES the whole set, so anything omitted stops being delivered. Changing the target URL re-checks the existing subscriptions against it, and a collision with another webhook is rejected with 409. The signing secret does not change and is not returned.
Attio: Upsert List Entry
Add a record to a list, or update its existing entry if it is already on there. This is the idempotent way to put something into a pipeline. Two things to know: multiselect attributes are made to match the values you send exactly, so anything omitted is removed, and if the record already has SEVERAL entries on this list Attio cannot choose between them and returns MULTIPLE_MATCH_RESULTS.
Attio: Upsert Record
Create a record, or update the existing one that matches. Attio looks for a record whose matching attribute holds the same value; if it finds one it updates it, otherwise it creates. This is the node for ingesting leads, form submissions and webhook payloads, because it is the one that does not duplicate on a retry. One subtlety worth knowing: if the MATCHING attribute is multiselect, supplied values are added and nothing is removed, but every OTHER multiselect attribute is made to match the supplied list exactly, which deletes values you did not send.
Connecting Attio takes one credential
Attio takes one static token in one header on every request, sent as an Authorization bearer header. Workspace API keys are created in Attio's workspace settings under the developer section, where you also choose the scopes the key carries, and Attio lets you edit those scopes afterwards rather than forcing a new key. Scopes are per family: object configuration, record permission, list configuration, list entry, user management, note, task, comment, email, meeting, call recording, file and webhook, each available read only or read and write. Give a key only the families your workflows need. Attio also accepts HTTP Basic with the token as the username, documented as a fallback, but bearer is the stated preference and the only form this connector uses. Two plan facts belong here rather than in a footnote. Attio's pricing comparison lists API and webhook access on every tier including Free, so a free workspace can create a key and run most of this catalog; confirm that on Attio's own pricing page before you build around it, since plan contents move. And two surfaces are gated: Attio's read only SQL endpoint states in its own specification that the workspace must be on the Enterprise plan, and creating or updating a list with advanced workspace access or per member access can be refused with a billing error on plans without it, though list creation with plain full access works. Separately, the email metadata endpoint is alpha and enabled per workspace on request to Attio's support team, and meetings, call recordings, files, record search and SQL are all documented as beta.
In Attio, open your workspace settings, go to the developer section, and create an access token. Choose the scopes your workflows need and copy the value.
In Mattering Automations, open Connectors, click Add Connector, choose Attio, and create a credential with that token.
Click Test Connection. Mattering calls Attio's token introspection endpoint, which is the only operation in Attio's whole specification that requires no scopes at all, so a narrow key still passes the connection test for the right reasons. On success it names the workspace the key belongs to and reports the scopes it carries, which is the fastest way to see whether the scopes are the ones you meant.
All 76 Attio nodes appear in the canvas palette. Run Attio: List Objects first. It returns every object in the workspace, both the ones Attio ships and the ones your workspace defined, and it is the node that fills the object picker every attribute and record node depends on.
Common uses for the Attio MCP
Ingest without creating duplicates
Attio: Upsert Record is the node most ingest paths want. It looks for a record whose matching attribute holds the same value, updates it if it finds one and creates it otherwise, which makes a retried workflow idempotent instead of dangerous. Attio: Create Record deliberately does not do this: it always creates, and throws on a conflict with a unique attribute rather than updating, which is the right behaviour when a duplicate should be an error. One asymmetry to know before you rely on upsert: for the MATCHING attribute, supplied multiselect values are added and nothing is removed, but for every OTHER multiselect attribute the supplied list becomes the entire list, so omitting a value removes it.
Move records through a pipeline from a workflow
In Attio a pipeline is a list, and a record's position in it is a list ENTRY carrying its own attributes: stage, owner, value, whatever the list defines. Attio: Upsert List Entry is the idempotent way to put a record into a pipeline, Attio: Update List Entry changes the entry's own attributes, and Attio: List Record Entries answers the question of which pipelines a company is currently in. Attio: List Statuses reads the stage vocabulary a list or object offers, which is worth running before writing a stage, since Attio matches on the option title. The distinction that saves data later: to take something out of a pipeline, delete the list ENTRY, not the record. Deleting the record removes it from every list and takes its whole value history with it.
Land AI research where a human will actually see it
Attio: Create Note writes onto a record and accepts markdown, so headings and lists survive, which makes it the natural landing place for a call summary or a research brief an agent produced. Attio: Create Task creates the follow up, assigns it and links it to records, and its linked records field is unusually forgiving: it accepts a record id, a company domain or a person's email address. Attio: Create Comment starts a thread on a record or a list entry, and one behaviour makes it more useful than it looks: putting a workspace member's email address in the comment content NOTIFIES that person, which is how a workflow pings an owner rather than filing something silently. It is also the reason not to paste arbitrary contact lists into a comment body.
Get calls and meetings into the CRM automatically
Attio: Create Meeting records a meeting with its participants and linked records, which is how an external calendar or notetaker syncs its events in, and Attio: Create Call Recording attaches a recording with its transcript, which is how a notetaker gets its output where the deal lives. Attio: Get Call Recording reads the transcript back as timed segments with speaker attribution plus a flat readable version, which is a good input for a summarisation step that then writes a note. Two behaviours to plan around, both documented by the vendor. A meeting's participants given by email address automatically create person and company records, and there is no built in de duplication, so a re run creates a second meeting rather than updating the first. Meetings, call recordings and files are all documented as beta.
Reconstruct history instead of guessing at it
Attio keeps value history rather than only the current value, and two nodes expose it. Attio: List Record Attribute Values and Attio: List Entry Attribute Values read one attribute's values on a record or an entry, optionally including every historic value oldest first, each carrying the window it was active for. That is how you reconstruct a deal's stage history, or answer when an owner actually changed, without an audit product. Their write siblings are the two most dangerous nodes in the connector and are flagged accordingly: replacing history destroys every value the attribute currently holds, including values not in your request, and Attio requires an explicit acknowledgement flag before it will do it.
Report on a workspace whose shape you do not control
Attio: List Objects, Attio: List Attributes, Attio: List All Lists and Attio: List Select Options are how a workflow discovers a workspace's actual schema rather than assuming one, which matters because non system slugs change when somebody renames a thing in the UI and Attio's own guidance is to store ids when an integration must survive that. For output that drops straight into a chart, Attio: Query SQL is the one node here that returns flat rows, and it is Enterprise only by Attio's own specification. On other plans, Attio: Search Records is the flattest read available, returning a display name and an object slug per row, and Attio: List Records is the strongly consistent one to use when a workflow has just written something and needs to read it back.
Example: An inbound lead that lands in the pipeline with its research already attached
Built on the Automations canvas in a few minutes, or described in plain language to Matter and built for you.
Trigger: a form submission or an inbound email arrives with a company name and a contact address.
A research step, an agent or a web node, gathers what is publicly known about the company.
Attio: Upsert Record writes the company on the matching attribute, so a lead that has arrived before updates the existing record instead of creating a second one. Attio: Upsert Record runs again for the person, matched on email address.
Attio: Create Note writes the research onto the company record as markdown, so the brief lives next to the record rather than in a chat transcript.
Attio: Upsert List Entry adds the company to the inbound pipeline list and sets the entry's own stage and owner attributes. Run Attio: List Statuses first so the stage title matches the vocabulary the list actually offers.
Attio: Create Task opens the follow up, assigned to the owner and linked to the record. If the owner should hear about it immediately, Attio: Create Comment starts a thread on the record with their email address in the content, which is what notifies them.
Nothing in this workflow can delete anything: it touches only upserts, creates and reads, so the 11 destructive nodes stay out of reach of an automated path.
Attio MCP questions, answered
What Attio operations are supported?
76 nodes on Attio's v2 REST API: 38 read nodes and 38 action nodes. Objects, attributes, select options and statuses, all with create and update, and archiving instead of deletion. Records with query, get, create, upsert, update, overwrite, merge and delete, plus attribute value reads and history replacement and the record entries lister. Fuzzy record search and read only SQL. Lists, list views, object views, and list entries with the same create, upsert, update, overwrite and delete family plus their own attribute value history. Workspace members, notes, tasks, comment threads and comments, email metadata, meetings, call recordings with transcripts, files and folders, and webhooks. Deliberately not included: Attio's multipart file upload and binary file download, which do not fit a JSON node, a deprecated transcript endpoint that the call recording node already covers, and Attio's SCIM identity provisioning surface, which is a separate API driven by an identity provider rather than by an automation.
Does this work with Claude, ChatGPT, or other AI assistants?
Yes. Mattering exposes these nodes over MCP, an open protocol for connecting AI assistants to tools, so any MCP capable assistant, including Claude and ChatGPT, can connect to your workspace and use them. Matter, the assistant built into the platform, uses the same nodes natively.
Attio ships its own MCP server. Why use this connector?
It does, and for some jobs its server is the better answer. Attio lists an MCP server across its plans, and if what you want is an assistant that can reach Attio during a conversation, that is a direct route with nothing else in between. This connector is shaped for operations rather than for conversation: the same 76 nodes are steps on an automation canvas as well as MCP tools, so they run on a schedule or from a trigger with nobody present, chain with every other connector in one workflow so an Attio record can be created from a form submission and trigger a message elsewhere in the same run, hold the credential once in a shared workspace rather than in each person's local client config, and carry destructive flags you can gate on. The honest summary is that they answer different questions: use Attio's server for an assistant that talks to Attio, and this for work that has to happen whether or not anyone is in a chat window.
Which nodes are plan gated or not generally available?
Three separate things, and it is worth keeping them apart. Plan gating: Attio: Query SQL states in Attio's own specification that the workspace must be on the Enterprise plan, and creating or updating a list with advanced workspace access or per member access can be refused with a billing error on plans without that feature, though list creation with plain full access works. Access gating: Attio: List Emails is documented as alpha and enabled per workspace and per app on request to Attio's support team, so a token that works everywhere else can be refused there. Maturity: Attio documents meetings, call recordings, files, record search and read only SQL as beta, which means they are live and callable but Attio reserves the right to make small changes. Attio's pricing comparison lists API and webhook access on every tier including Free, so the base of this catalog is not plan gated at all; confirm the current split on Attio's pricing page before you design around it.
Can an AI agent delete my CRM data by accident?
The 11 destructive nodes are flagged so you can gate them, and Attio's deletes are genuinely final: its own archiving guide says deleted data is eventually removed entirely with no ability to restore it, so there is no trash and no restore window to fall back on. The ones worth reading twice are Attio: Delete Record, which takes every attribute value, its history, and the record's entries in every list; Attio: Merge Records, which destroys the secondary record with no unmerge; Attio: Replace Record Attribute History and Attio: Replace Entry Attribute History, which destroy an attribute's whole value history including values you did not send; Attio: Delete Comment, which takes the entire thread when the target is the first comment in it; and Attio: Delete File, which takes every descendant when the target is a folder. Keep agents on the read nodes, the create and upsert nodes and the update nodes, and put the deletes behind a condition or an approval task. Where Attio offers a reversible alternative it is usually the right one: archive an attribute or an option rather than expecting a delete, mark a task completed rather than deleting it, and remove a list ENTRY rather than the record when the intent is to leave a pipeline.
Why are update and overwrite separate nodes?
Because the difference is data loss rather than style, and hiding it behind a checkbox would be the wrong call. On records and entries, Attio's PATCH APPENDS to multiselect attributes: values you send are added to what is already there and nothing is removed. Its PUT OVERWRITES them: for every multiselect attribute you mention, the supplied list becomes the entire list and anything not in it is removed. So Attio: Update Record and Attio: Update List Entry append, Attio: Overwrite Record Values and Attio: Overwrite List Entry Values replace, and each node says which it is in its own description. Attributes you do not mention are untouched either way.
How does the connector cope with a CRM that has no fixed schema?
By discovering it at run time rather than assuming it. Attio's objects, attributes, lists and entries are all workspace defined, so the nodes carry dependent pickers instead of hard coded field lists: Attio: List Objects fills every object field, the chosen object scopes Attio: List Attributes so you only see attributes that exist on it, and the chosen attribute scopes the select option and status pickers below that. Lists work the same way down to entries. One consequence worth knowing when you template a workflow: Attio's system object and attribute slugs are stable, but non system slugs change when somebody renames the thing in the UI, and Attio recommends storing ids for integrations that must survive that. Every endpoint accepts either, so a long lived automation against a custom object can use the id.
Why can I not read a record's name from the record lister?
Because of the shape Attio returns it in, and this is the one thing most likely to cost a first time builder an hour. A record's values arrive as a MAP keyed by attribute slug, each mapping to an array of value objects, rather than as a flat row. There is no fan out path across values, so a person's name reads as a specific path through the map rather than through a wildcard, and a picker built on the lister would have to label every option with its own identifier. That is why record pickers here read from Attio: Search Records instead, which returns the display name Attio itself renders plus the object slug, and takes a genuine free text query. The trade is stated on every field that uses it: search is beta and eventually consistent, so a record created seconds ago may not appear yet. When a workflow has just written something and must read it straight back, use Attio: List Records, which is strongly consistent.
What are Attio's rate limits?
Attio documents 100 read requests per second and 25 write requests per second, not documented as varying by plan, and a rejected request returns HTTP 429 with a retry hint. A rejected request was not executed, so it is safe to retry, usually on the next second. Two specifics matter more than the headline numbers. Creating a call recording is capped at one request per second, far below the general write limit, so throttle any backfill of historical calls. And Attio: List Records and Attio: List Entries additionally score query complexity: a single over complex query is rejected and has to be simplified, and cumulative score across a short sliding window can also trip the limit, which is worth knowing before pointing a dashboard at a heavily filtered list query on a fast refresh.
Put Attio to work inside your workspace
Connect once, then use every node in visual automations and AI agent workflows across the Mattering platform.
All product names, logos, and brands are property of their respective owners and are used for identification purposes only. Use of these names, logos, and brands does not imply endorsement or affiliation. Attio is a product of its respective owner; Mattering is an independent platform.