Find which Attio accounts an outage touched, from a Sazabi thread

Last verified on Nov 8, 2026.

When something breaks in production, the first question from the business side is who it hit. Your logs know which tenants, domains and user IDs produced errors. Your CRM knows which of them are customers, which are mid-trial, which have an open renewal and who owns the relationship. Attio holds that second half, and the Sazabi agent can read it in the same thread where it reads the first.

What the Sazabi agent can do with Attio

Sazabi connects to Attio's hosted MCP server at mcp.attio.com/mcp over OAuth, and you choose the Attio workspace during authorization. Attio groups its tools by job:

• Records and objects: search-records does full-text search across people, companies and deals by name, email or domain. list-records lists one object with filters and sorting, and get-records-by-ids pulls complete profiles. list-attribute-definitions shows which fields an object has and what its options are.
• Lists: list-lists finds your pipelines, and list-records-in-list returns entries with their parent records.
• Notes and emails: semantic-search-notes finds notes by topic, get-note-body reads one, and search-emails-by-metadata finds email by participants, domain or time range.
• Tasks and comments: list-tasks reads open work, and create-task and create-comment write follow-ups.
• Reporting: run-basic-report counts and groups records or list entries, and query-particle-sql runs read-only SQL, which Attio says is not on every billing plan.

Attio's docs say read operations are auto-approved while write operations ask for confirmation. Sazabi adds its own control: write tools need explicit approval under your organization's tool policy, and --read-only on sazabi mcp-connectors connect attio restricts the connector to read tools.

Questions it answers

• Which of the domains in these error logs are customers, and who owns each account?
• Are any of them in our enterprise list, and at what stage?
• Is there an open task or recent note on this account that explains what they were doing?
• How many list entries sit in each stage for the accounts we think were affected?
• Who on the team should be told?

A worked example: ranking an outage by account

A queue worker failed for a few hours. Your logs show failed jobs grouped by the tenant's email domain. You ask:

From the log search, take the ten tenant domains with the most failed jobs
between 11:00 and 14:00 UTC. Look each one up in Attio. For every match, tell me
the account owner and which list it is in. Then list any open tasks linked to
those companies.

The agent runs the log query first and shows it. It then calls search-records with each domain, which matches company records by domain. For hits it calls list-records-in-list against the list you name, and get-records-by-ids for owner and stage fields. list-tasks filtered to those companies shows what is already in flight.

The result is a prioritized list, but it is only as good as your CRM. A domain with no Attio match is not necessarily unimportant. It may be a personal address, a subsidiary or a record that was never created, so the agent should say which domains it could not match. Once you have the list, ask it to draft a create-comment on each affected record, or a create-task for the account owner. Those are writes, so they wait for your approval.

How it fits with other integrations

CRM context pairs well with the tools where the response happens. Slack is where the thread usually lives, Linear tracks the fix, and Intercom or Pylon show what customers have written in. For the wider catalog see integrations, and for plans see pricing.