Connect Pylon issues and accounts to what your logs show

Last verified on Nov 8, 2026.

Support tickets are a record of how a failure looks from outside. Customers give you the time, the account, the exact words of the error, and sometimes a screenshot. Your logs show the same event from inside. Pylon holds the first half, and the Sazabi agent can read it in the same thread where it reads the second, so you can see which customers hit an error and which of them already have an engineering ticket.

What the Sazabi agent can do with Pylon

Sazabi authorizes with Pylon over OAuth. Pylon says the connection acts for the authenticated user and sees only what that user sees in the dashboard. Member and Admin users need the MCP Access role. Pylon's tool reference marks each tool read or write. For an investigation the read tools matter most:

• search_issues searches by keyword and filters by state, account, assignee, tags and custom fields. Each result lists its linked engineering tickets and their state. With include_messages it returns the conversation, and max_messages_per_issue and max_message_chars trim large pages.
• get_issue and get_issue_messages read one issue and its full message history, including the original To and Cc recipients of email messages.
• search_accounts and get_account find accounts by name, domain or external ID and return custom fields. get_account_notes reads account notes where your plan includes Account Intelligence.
• search_contacts and get_contact look up people by name, email, phone or account.
• get_issue_threads and get_issue_followers show internal threads and followers.

The write tools are the part to be deliberate about. send_issue_note posts an internal note visible only to your team. send_issue_reply goes straight to the customer on the issue's channel and cannot be unsent. link_product_issue attaches a Linear, Jira, GitHub, Asana or Shortcut ticket. Sazabi says write operations are restricted by default, and --read-only on sazabi mcp-connectors connect pylon keeps the connector to read tools.

Questions it answers

• Which open issues mention this error message, and from which accounts?
• Do those issues already have a linked engineering ticket, and what state is it in?
• Who is the contact on the account, and what did they write?
• Are several accounts reporting the same thing in the last hour?
• Who owns the issue, and which team is it assigned to?

A worked example: counting the customers behind one error

Your logs show a spike of invoice_pdf_render_failed. You ask:

Search Pylon for open issues that mention "invoice PDF" opened in the last 6
hours. For each, give me the account, the contact, the assignee and any linked
engineering ticket with its state. Do not post anything.

The agent runs the log query and shows it. It then calls search_issues with the keywords and the open state, and reads each result's linked engineering tickets from the response. For accounts and contacts it calls get_account and get_contact, and for the wording it asks for the messages and trims long bodies with max_message_chars. Then it lines up the issue timestamps against the first error in your logs.

You get a count of affected accounts and a note on coverage: which issues already have a ticket and which do not. That second list is the useful one. If six issues describe the failure and none link to engineering, the fix has no owner yet. When you are ready, ask for a send_issue_note on each, saying the issue is under investigation, or for link_product_issue against the ticket your team opened. Both are writes, so check what Sazabi's connection allows first. Leave send_issue_reply to the support team, because a customer message cannot be recalled.

How it fits with other integrations

Pylon issues link out to the trackers where engineering works, so pair it with Linear or GitHub. Slack hosts the thread, and Intercom covers teams whose support runs there. Browse all connectors under integrations, or see pricing.