Slack ticket reactions turned into Linear issues by AI

Slack ticket reactions turned into Linear issues by AI

An n8n workflow that watches one Slack channel for messages marked with a ticket reaction. For each message that has no
issue yet, an AI model writes a title, a summary, a few ideas and a priority, and the workflow creates the issue in
Linear. The Slack message id is saved in the issue, so no message is ticketed twice.

What is inside

• workflow.json: the workflow as n8n exports it, 12 nodes and 7 sticky notes. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• The model node named no model, so n8n would run its own default for that node version, gpt-5-mini. OpenAI's
  deprecations page, read on 12 November 2026, shuts it down on 11 December 2026. The node now names gpt-6-luna. It
  sits under a plain chain with a structured output parser and calls no tool, so that model fits. The chain is now called
  Generate Ticket Using AI.
• Get Hashes Only looked for a hash: line in every issue Linear returned and took the first match by position, [1]. An issue
  with a description but no such line, such as one written by hand, made the match return null and stopped the run. So did
  a Linear workspace where no issue has a description, because the list was then missing. The expression now skips both.
• The message id came from the last part of the Slack link. Slack's documentation shows links to a reply in a thread
  ending in a query string, ?thread_ts=...&cid=.... The saved hash: line is read back only up to that character, so
  such an id would never match and the same reply could be ticketed again at every run. The id now ignores the query
  string.
• The message was placed into JSON by hand, with only quotes and line breaks escaped. A backslash or a tab in the text, as
  in pasted code, could break it. JSON.stringify now does the escaping.
• The Linear team id and the state id belonged to the original author's workspace. The team is blank, so you choose
  yours, and the state is left out, so Linear chooses it.
• Get Existing Issues now runs once per run. It had run once per message, and each run read the same issues.
• The search channel, #n8n-tickets, is now #support-tickets. The notes on the canvas had links and an invitation to a
  community. They are rewritten, and one note now states the 5 minute schedule.

Before you start

• An n8n instance that stays switched on, on n8n Cloud or on your own server. It asks Slack and Linear for news, so it
  needs no public address.
• A Slack user token with the search:read scope. Slack's message search takes a user token, and a bot token cannot use
  it.
• An OpenAI API key.
• A Linear API key, and the team that should get the issues.

Set it up

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add a Slack API credential with your user token to Slack. In its query, replace support-tickets with your channel.
3. Add an OpenAI credential to OpenAI Chat Model.
4. Add a Linear credential to Get Existing Issues and to Create Ticket, then choose your team in Create Ticket.
5. In the channel, react to a message with :ticket: and run the workflow once with Execute workflow. Look for the new
   issue in Linear.
6. Switch the workflow on. It then runs every 5 minutes.

How a run goes

1. Schedule Trigger starts the run, and Slack searches for the newest 10 messages in the channel that carry a ticket
   reaction.
2. Get Values gives each message an id taken from its link, and keeps the channel, the author, the time, the link and the
   text.
3. Get Existing Issues reads issues from Linear. Collect Descriptions gathers their descriptions, and Get Hashes Only
   pulls the saved ids out of them.
4. Merge pairs every message with that list. Create New Ticket? lets a message through only when its id is not in it.
5. Generate Ticket Using AI asks the model for a short title, a summary, up to 3 ideas and a priority.
   Structured Output Parser turns the answer into those fields.
6. Create Ticket creates the issue. Its description holds the summary, the ideas, the original message with its author,
   the channel, the time, the link and the hash: line. The priority maps urgent to 1, high to 2, medium to 3 and low to 4.

Change the look

• The channel or the reaction: the query in Slack.
• How often it looks: Schedule Trigger.
• How many messages one run reads: the limit in Slack.
• What the ticket says: the prompt in Generate Ticket Using AI.
• The model: OpenAI Chat Model.
• The team, labels or priority mapping: Create Ticket.

Good to know

• The facts below come from Slack, OpenAI, Linear and n8n pages and source read on 12 November 2026. Nothing was run: no
  Slack search, no model call and no Linear issue.
• The search runs as the person whose token you use, so it sees what that person can see. Any member who can react in
  the channel can open an issue.
• Slack's search returns 10 messages here, newest first by n8n's default. If more reacted messages than that are newer
  than one you wanted, raise the limit in Slack.
• Linear is read with n8n's default limit of 50 issues, across all teams. If the issue for a message is not among them,
  the message can get a second issue. Raise the limit in Get Existing Issues, or turn on Return All.
• The priority is the model's guess from the tone of the message, and the ideas are not checked.
• gpt-6-luna costs USD 0.1 per million input tokens and USD 0.5 per million output tokens on OpenAI's page. A page's text
  can change, so check it before a large run.
• Your data: the message text goes to OpenAI. The message, the author's name and id, the channel and a link go to Linear.
• Credentials: the JSON holds none.