How to Build an AI Agent for Linear and Slack

The workflow problem

Bug reports, feature requests, and operational issues surface in Slack every day. A customer support rep pastes an error screenshot in #support-escalations. A product manager describes a feature idea in #product-feedback. An engineer reports a flaky test in #engineering. Each of these messages represents work that should be tracked in Linear, but the path from Slack message to Linear issue is paved with friction.

The current process looks like this: someone reads the Slack message, opens Linear in a new tab, creates an issue, tries to capture the context from the Slack conversation in the issue description, assigns a priority based on gut feeling, puts it in the right team’s queue, then goes back to Slack and replies with the issue link. This takes three to five minutes per issue. More importantly, it requires a human to be actively monitoring channels and making triage decisions.

Many issues never make it to Linear at all. During busy periods, Slack messages get buried under new conversations. Support escalations sit unread over weekends. Feature requests from customers get a thumbs-up emoji but no issue created. The team loses track of commitments made in chat, and the same issues get reported multiple times because there is no central record. Sprint planning is based on whatever issues someone remembered to create, not a comprehensive view of incoming work.

Why an AI agent, not just automation

Slack-to-Linear integrations exist as bots that create issues when someone uses a slash command or adds a specific emoji reaction. These tools require a human to decide that an issue should be created and to manually input the title, description, and priority. They reduce the tab-switching but not the judgment overhead.

An AI agent eliminates the triage bottleneck entirely. It monitors designated Slack channels and identifies messages that represent actionable work, distinguishing between casual conversation and genuine bug reports or feature requests. When it detects an issue-worthy message, it does not just transcribe it. It reads the full conversation thread, extracts the core problem, identifies affected systems or customers, and assigns a priority score based on factors like customer tier, system criticality, and how many people reported the same problem.

The agent’s scoring model goes beyond simple keyword matching. A message saying “the app crashed” in #random from someone testing a dev build is different from the same message in #support-enterprise from a rep handling a $200K customer. The agent considers the channel (support versus general), the reporter’s role (support rep versus intern), mentions of specific customers or systems, and the urgency language used. This contextual triage is something no static automation can provide.

The agent also prevents duplicates. Before creating a new issue, it searches Linear for existing issues with similar descriptions. If it finds a potential duplicate, it adds the new context as a comment on the existing issue and notifies the Slack reporter that the issue is already being tracked, along with its current status.

How it works with ACP

ACP provides your agent with access to Slack’s message stream and Linear’s project management system, enabling it to turn conversations into tracked, triaged work items.

Slack tools available to the agent:

• Read messages in channels the agent is invited to
• Read full conversation threads for context
• Post messages and threaded replies
• React to messages with emoji
• Read user profiles for role and team context
• Access channel metadata (name, purpose, topic)
• Send direct messages for triage confirmations

Linear tools available to the agent:

• Create issues with title, description, priority, labels, and team assignment
• Search existing issues by text, label, and status to detect duplicates
• Read issue details and comments
• Update issue status and priority
• Assign issues to team members
• Add comments to existing issues
• Read team structures and workflow states
• Access cycle/sprint information for scheduling

The agent turns Slack into an intelligent intake system for Linear. Here is a realistic scenario:

Slack message in #support-escalations from Sarah (Enterprise Support Lead): “Just got off a call with Pinnacle Financial. Their admin dashboard has been showing incorrect user counts since yesterday. They have 850 users but the dashboard shows 412. This is blocking their quarterly compliance audit. They need this fixed urgently.”

Agent analysis: “This is from #support-escalations, posted by an Enterprise Support Lead. It describes a data accuracy bug affecting Pinnacle Financial, an enterprise customer. The issue involves incorrect user counts in the admin dashboard. The phrase ‘blocking compliance audit’ and ‘need this fixed urgently’ indicate high urgency. Let me check for existing issues.”

Duplicate check: The agent searches Linear for issues mentioning “admin dashboard,” “user count,” or “Pinnacle Financial.” It finds no matching open issues.

Triage score: Priority 1 (Urgent). Factors: enterprise customer, data accuracy issue, compliance impact, explicitly reported as urgent by senior support staff.

Linear issue created:

• Title: “Admin dashboard showing incorrect user counts for Pinnacle Financial”
• Description: “Enterprise customer Pinnacle Financial reports their admin dashboard displays 412 users instead of the actual 850. Issue started yesterday. This is blocking their quarterly compliance audit. Reported by Sarah (Enterprise Support) via Slack on June 15.\n\nAffected: Admin dashboard user count module\nCustomer: Pinnacle Financial (Enterprise tier)\nImpact: Compliance audit blocked”
• Priority: Urgent
• Team: Platform
• Labels: bug, enterprise, customer-reported

Slack response (threaded reply): “Created Linear issue ENG-891: ‘Admin dashboard showing incorrect user counts for Pinnacle Financial’ – Priority: Urgent, assigned to Platform team. View in Linear”

Governance and security

Converting chat messages into tracked work items involves handling conversational data that may contain sensitive information. ACP ensures appropriate boundaries.

Slack OAuth scopes: The agent uses channels:history and groups:history to read messages in channels it is added to, and chat:write to post responses. It does not have access to direct messages between other users, private channels it hasn’t been invited to, or workspace admin functions. The agent cannot modify channel settings, manage users, or access Slack Connect shared channels unless explicitly configured. You control exactly which channels the agent monitors by inviting it to specific channels.

Linear OAuth scopes: The agent authenticates with Linear’s OAuth flow, receiving permissions to create and update issues, read project data, and manage comments. It cannot modify workspace settings, manage team memberships, or delete issues. Issue creation is limited to designated teams, so the agent cannot accidentally create issues in the wrong team’s board.

Message handling: The agent processes Slack messages in real time but does not store conversation content beyond what is written into Linear issue descriptions. Slack messages that the agent determines are not actionable (casual conversation, off-topic discussion) are not logged or retained. You can configure sensitivity filters to prevent the agent from including certain types of information (personal data, salary discussions, legal matters) in Linear issues.

Audit trail: Every issue creation is logged with the originating Slack message ID, channel, reporter, and the triage reasoning the agent applied. If a stakeholder questions why an issue was created with a particular priority, the audit trail shows exactly what information the agent used to make that decision. This transparency builds trust in the automated triage process.

Human override: The agent can be configured to require confirmation before creating high-priority issues. In this mode, it posts a preview in Slack with a “Create this issue?” prompt, and a team member confirms before the Linear issue is actually created.

Example use cases

• Bug report intake: Support reps paste error details and customer context in a dedicated Slack channel. The agent creates Linear issues with structured descriptions, appropriate priority based on customer tier and severity, and assigns them to the correct engineering team.

• Feature request collection: Product managers and customer-facing teams share feature ideas in Slack. The agent creates Linear issues labeled as feature requests, checks for similar existing requests to prevent duplicates, and adds vote counts when multiple people request the same thing.

• Incident-to-issue conversion: During incident response, the on-call engineer works in a Slack channel. After the incident is resolved, the agent reviews the conversation thread and creates follow-up Linear issues for root cause investigation, monitoring improvements, and documentation updates.

• Cross-team request routing: When a message in a design team’s Slack channel describes work that requires engineering effort, the agent creates the issue in the engineering team’s Linear board while keeping a link back to the original design discussion for context.

• Standup and status queries: Team members ask the agent in Slack “What are my open issues?” or “What’s the status of ENG-891?” and receive formatted responses without switching to Linear. This keeps daily standups flowing in Slack with live data from the project tracker.

Getting started

Build your Slack-to-Linear triage agent quickly:

1. Sign up at cloud.agenticcontrolplane.com. Free tier includes everything needed.

2. Connect Slack and Linear. Authorize both tools on the Data Sources page via OAuth. Invite the agent to the Slack channels you want it to monitor and select the Linear teams it can create issues for.

3. Describe your agent. Tell ACP: “Monitor #support-escalations and #engineering-bugs for actionable issues. Create Linear issues with appropriate priority based on customer impact and severity. Check for duplicates before creating. Post the issue link as a threaded reply in Slack.” Your intelligent intake system is running.