Connect Claude to Attio

Query the records on one object - this is the main way to read anything out of Attio. `object` is the object's slug or id, from list_objects: "people", "companies" and "deals" exist in most workspaces, but a workspace defines its own objects, so check rather than assume. `filter` is Attio's own filter document keyed by attribute slug, and `sorts` is a list of {"direction": "asc"|"desc", "attribute": "<slug>"} entries. Call list_attributes for the attribute slugs this workspace actually has before filtering on anything beyond the obvious ones - a filter naming an attribute that does not exist comes back as an error rather than as an empty result. Pages by offset, not by cursor.

Fetch one record in full, with every attribute value Attio holds for it. Takes the object slug and the record's own id, both of which query_records returns. Use this after a query when you need the values the query did not show you.

List every object this workspace defines, standard and custom, with its slug, singular and plural names and id. Call this first whenever you do not already know what the workspace calls something: Attio ships People, Companies, Deals and Users by default, and a workspace is free to add its own or rename those. The slug is what every other operation here takes as `object`.

List the attributes defined on one object, with each attribute's slug, title, type and whether it is unique. This is the translation table between what a person says and what the API takes - "the company's website" is an attribute with a slug, and query_records, create_record and update_record all speak in slugs. Call it before filtering or writing anything you have not seen this workspace return.

List the workspace's lists - Attio's pipelines and saved collections, each with its slug, name, id and which object its entries point at. A list is where a workspace tracks a process (a sales pipeline, an onboarding queue), so this is the operation that finds the pipeline before list_list_entries reads what is in it.

Read the entries on one list - the records currently in that pipeline or collection, with the per-entry attribute values the list itself holds (stage, owner, dates). Takes the list slug or id from list_lists. `filter` and `sorts` have exactly the same shape as query_records, but the attribute slugs are the list's own, which list_attributes returns when asked about a list rather than an object.

List notes, newest first - the written history of what people in the workspace have said about a record. Give `parent_object` and `parent_record_id` together to read one record's notes; leave both off to read the workspace's recent notes across every record, which is a much broader read and should be paged rather than pulled whole.

List the people who have access to this Attio workspace, with each member's id, name, email address and access level. Call it to turn a colleague's name into the workspace member id that an owner or assignee attribute takes. These are the caller's own colleagues, not records about anybody the workspace sells to.

List tasks, with their text, deadline, assignees and the records they are linked to. Narrow it with `linked_object` plus `linked_record_id` to see one record's outstanding work, or with `assignee` to see one person's. `sort` takes Attio's own combined field and direction, e.g. "created_at:desc".

Report what this connection's API key actually is: which workspace it belongs to, which workspace member created it, and the exact list of scopes it was granted. It needs no scope of its own, which is why it is the connection check - it answers "is this key real and whose is it" without depending on any permission the key may not have been given. Call it when a write comes back refused, to see whether the scope is missing rather than the record.

Create a record on one object. `values` is a map of attribute slug to value, in Attio's own value shape - a text attribute takes a string, a multiselect takes a list, and a record-reference attribute takes a list of {"target_object", "target_record_id"}. Call list_attributes first for the slugs and types this workspace defines. Attio refuses a create that collides with a unique attribute, so query_records first when the record may already exist and use update_record when it does.

Change attribute values on an existing record. Only the attributes you send are touched. This uses Attio's appending update - a value sent for a multiselect attribute is added to what is already there rather than replacing it, which is the reversible half of the two update verbs Attio offers; the overwriting one is deliberately not exposed. This cannot delete the record.

Attach a note to a record. The note appears on that record's timeline inside the workspace; it is not sent to anybody outside it, which is why this is a write rather than a destructive action. Say which object the record belongs to and give its id - both come from query_records. `format` is "plaintext" or "markdown"; markdown supports headings, lists, emphasis and links, and image embeds are not accepted through the API. `created_at` back-dates the note and may not be in the future.