Loops

Configured Linear Agent workflows that execute in the background from a time schedule or a supported Linear workspace event. Each execution starts its own Linear Agent conversation, gathers applicable context, performs supported actions within the Loop's configured scope and permissions, and is recorded in the Loop's run history. A completed run can subsequently be continued as an ordinary Agent conversation. Linear publicly introduced Loops on 20 July 2026; Linear does not apply a preview, beta or general-availability label to Loops in the primary material reviewed.

Recorded characteristics

Function

Linear documents Loops as a way to "let Linear drive work forward automatically": they run on a set schedule or when supported triggering events occur on issues, projects, initiatives, cycles or releases, follow written instructions, and use context from the workspace and connected tools to decide on the next best action. Execution model as documented: a Loop is configured (trigger, instructions, optional MCP connectors, scope and permissions) and published; when its schedule fires or a supported event occurs and the configured conditions match, the Loop runs; each run starts its own conversation with Linear Agent, retrieves applicable context, selects and performs supported actions within the Loop's configured scope, and is recorded in the Loop's run history; once a run finishes, a person can continue that conversation like any other agent session to review the work, add context, or direct what happens next. Recurring runs are separate conversations, not one persistent conversation. Documented schedule triggers: hourly, daily, weekly (one or more selected days of the week), monthly and yearly, with a chosen timezone and a chosen first-run date that cannot be in the past. Documented event triggers on issues: creation, and changes to status, priority, assignee, cycle, project or labels. Documented project and initiative triggers (14 September 2026 expansion): changes to status, ownership, target dates, milestones, membership and labels, plus new updates and comments. Documented cycle triggers: cycles being created, started or completed. Triage: triage Loops are created for a specific team, require triage to be enabled for that team, run when an issue enters that team's triage queue and begins matching the configured conditions, and do not support schedules or non-issue triggers. Releases: the overview names releases among triggering event sources, but the primary documentation does not establish release trigger semantics; this is recorded as unresolved rather than described. Conditions: a Loop can include up to eight conditions; the configured conditions must be met before the Loop runs, and the Loop will not run again on subsequent updates while those same conditions remain matched. A Loop can include up to eight actions. Run now is documented as an additional manual trigger for enabled Loops that use Linear Agent and have a schedule, issue, project or initiative trigger (choosing the entity for issue, project and initiative triggers); Linear states Run now is not available for cycle triggers or for Loops that only apply configured issue updates. The availability of Run now does not change the normal scheduled or event-triggered execution model.

Data access

Linear documents that Loops use context from the workspace and connected tools. Access is bounded by the Loop's configured permissions: Team access controls which teams the Loop can access, and the Loop can read and write data only in those teams — by default a workspace-level Loop or a Loop in a public team has access to all public teams, while a Loop inside a private team has access only to that team, and Loops that can access all public teams can also access workspace-level objects such as Initiatives and Customers. Web access, if enabled, lets the Loop query any website, with Linear's documented caution that web access can send workspace content to external services. Code Intelligence, if enabled, lets the Loop browse and analyse repositories configured in the workspace. Externally synced issues and comments, if enabled, lets the Loop write on issues or comment threads synced with external applications. External sources: for security reasons Loops run by default only on issues created from within Linear; specific external sources (for example Slack messages or emails to an intake address) can be enabled per Loop from a list configured by workspace owners. Through the Slack connector, Loops can find public channels by name including Slack Connect channels and read their message history, and read messages from Slack links. Through MCP connectors, Linear documents searching or retrieving content from connected sources such as GitHub, Notion or Sentry, and fetching documentation, pull request details or error reports. Linear Agent itself is documented as working with existing workspace data — teams and sub-teams, initiatives and projects with milestones and cycles, issues and their relationships, comments, activity history and documents.

Actions

Can take actions

External actions

Yes

Human confirmation

Not required

Permission basis

Mixed

Administrative control

Documented controls. Setup: workspace owners manage who can create and manage workspace Loops in Settings → Security → Workspace management → Manage loops; team owners control who can create and manage Loops in Settings → Teams → [Team name] → Access and permissions → Loop management; per-Loop, "Who can edit" offers all members, workspace admins or team owners, or only the loop owner. Lifecycle: a Loop is created as a draft, edits are saved as a draft and applied only on Publish; all published versions are saved and can be restored, with the documented note that removed connectors cannot be restored because they must be authenticated again manually; Loops can be enabled or disabled by right-clicking the Loop in the workspace or team Loop view; deleting a Loop is permanent and cannot be undone. Permission toggles per Loop: Team access, Web access, Code Intelligence, Coding sessions, Externally synced issues and comments, External sources, and "Allow changes outside of triggering entity" — when that last setting is disabled the Loop can only write data on the entity that triggered the run, and Linear states it does not apply to scheduled Loops. Connector governance: Owners (Enterprise) or workspace Admins (Business) can allow users to connect any MCP connector or only allowlisted servers, and workspace admins and owners configure the full set of allowed MCP connectors in Security settings; Code Intelligence and Coding sessions each require a workspace admin to enable the feature and configure code access before the per-Loop permission can be used; Slack requires a Linear admin to turn on Enable Linear Agent and Allow Loops access in Settings → Integrations → Slack → Workspace. AI usage: workspace admins purchase and manage AI credits in Settings and review usage under AI & Agents > AI usage & credits. Documented limits: up to 100 Loops per team, up to 500 workspace-level Loops per workspace, up to 100 unpublished Loop drafts per person, up to 8 conditions and 8 actions per Loop, Loop names up to 64 characters and descriptions up to 255 characters. Loops run at the team and workspace level with shared visibility and control: anyone with access can review their instructions, see how they are configured, and inspect what happened during each run. Linear documents a loop run failed notification, with run history used to check recent runs and visible failures.

Default state

Conditional

Availability

Linear states Loops are "Available on paid plans". Loops were publicly introduced on 20 July 2026, at which point Linear stated they were available on Business and Enterprise plans. A material expansion documented on 14 September 2026 added project, initiative and cycle triggers and broader issue triggers, Linear document editing, drafting and posting Slack updates, and the ability to continue a completed run as an ordinary Agent conversation. Linear does not label Loops preview, beta or generally available in the primary material reviewed, and no lifecycle label is inferred here from public availability. Loops depend on Linear Agent, which Linear documents as available by default in a workspace and able to be disabled by a workspace admin or owner in Settings → AI → Linear Agent. A Loop does not exist until it is explicitly created and published, and its permissions, integrations and connectors must be configured for the actions it performs.

Licensing

Linear states that Loops use AI credits, which workspace admins can purchase and manage in Settings, that purchased funds expire 12 months after the purchase date, and that if a workspace runs out of credits Loops pause unless an admin purchases more credits or enables auto top-ups, with automatic charging only when auto top-ups are enabled. Linear states that Loops that use Linear Agent stop running when the workspace runs out of AI credits, while Loops that only apply configured issue updates can continue to run without AI credits. Promotional credits are recorded as temporary promotions and not as the price of Loops: at launch on 20 July 2026 Linear gave workspaces $20 per seat in promotional credits expiring 20 August 2026, and on 14 September 2026 Linear stated it had extended introductory Loops credits through 31 December. The Loops documentation also notes that some workspaces may receive promotional Loops AI credits with separate expiration terms, visible in the Usage and credits dashboard.

External model or provider

Not established

Limitations and uncertainty

Recorded boundaries and unresolved points. (1) Human confirmation: current Linear documentation describes background Loop execution after configuration and documents no per-run or per-action human approval or confirmation mechanism; oversight is provided through advance configuration and permission scope and through subsequent run inspection. Linear does not explicitly guarantee that no action ever requires approval, and run-time confirmation semantics are not specified — this is recorded as an open documentation point rather than settled. Creating and configuring a Loop is not treated here as contemporaneous human confirmation of each future action. (2) The in-Linear actor identity shown on issues, comments and document edits made by a Loop is not established by current documentation. (3) Document-edit attribution for Loop edits is not documented. (4) Document-level version history and restore or revert behaviour for Loop document edits is not documented; Linear's documented "Published versions" and "Restore version" behaviour applies to the Loop configuration, not to edited documents. (5) Model and provider: current primary documentation does not establish a specific model or provider for Loops or for Linear Agent, and models documented for coding sessions are not imported here. (6) The effect of disabling Linear Agent on already configured Loops is not documented. (7) Release trigger semantics are not established: releases are named among triggering event sources but no release trigger behaviour is documented. (8) Run history records when a Loop executed, what actions it took, and visible failures with common causes such as missing permissions, untrusted external sources or exhausted AI credits, and Slack messages posted by a Loop carry a direct link back to their run; Linear calls this run history for auditing behaviour, but tool-call-level granularity is not established and it is not described here as a formal audit log. (9) External capability is not unrestricted: Slack and MCP actions depend on the applicable integration or connector being connected, enabled for Linear Agent, allowlisted by administrators where applicable, and selected and permitted for the Loop, with an MCP connector using the configuring person's authenticated personal account. (10) Coding session boundary: a documented Loop action is starting a coding session, which Linear documents as able to open a draft pull request; the coding agent's subsequent repository edits, commits, pushes, tests and pull-request implementation work are not attributed to the Loop and are outside this record. (11) Slack restrictions are documented and recorded rather than generalised: Loops cannot access private channels, direct-message history or group direct messages, and cannot search messages across the workspace, create channels, edit or delete messages, upload files or manage Slack canvas documents; replies in Slack Connect channels, direct messages and unrelated threads do not continue a conversation; where automatic channel joining is disabled, Loops can only use public channels Linear has already joined and may be unable to read history there. (12) Autonomy is described as a scheduled or event-triggered unattended background agent workflow that is autonomous after configuration within its configured permissions; no stronger autonomy claim is made. (13) Generated analysis, recommendations and drafts are distinguished from persisted actions throughout this record.

Evidence

• Loops

  Supports: Function · Data access · Actions · External actions · Human confirmation · Permission basis · Admin controls · Default state · Availability · Licensing · Limitations · Primary source

  States Loops let Linear drive work forward automatically, running on a set schedule or when supported triggering events occur on issues, projects, initiatives, cycles or releases, following written instructions and using workspace and connected-tool context to decide on the next best action. Documents the configuration path (choose a trigger, add instructions, optionally add MCP connectors, review scope and permissions, create and publish), and that each Loop has a run history recording when it executed and what actions it took, which can be opened to ask Linear Agent follow-up questions or request additional actions using the run's context. FAQ states scheduled Loops can run hourly, daily, weekly, monthly or yearly with a chosen timezone and a first run date that cannot be in the past, and that weekly schedules can run on one or more days of the week. FAQ states triage Loops are created for a specific team, require triage to be enabled for that team, run when an issue enters the team's triage queue and begins matching the configured conditions, and do not support schedules or non-issue triggers. FAQ states that for Loops with defined conditions those conditions must be met before the Loop runs and that it will not run again on subsequent updates while the same conditions remain matched, with limits of eight conditions and eight actions. FAQ states Run now is available for enabled Loops that use Linear Agent and have a schedule, issue, project or initiative trigger, with entity selection for the latter three, and is not available for cycle triggers or Loops that only apply configured issue updates.

  Team access permission states the Loop can read and write data only in the teams it can access, with workspace-level and public-team Loops defaulting to all public teams and private-team Loops limited to that team, and workspace objects such as Initiatives and Customers reachable only by Loops with all-public-team access. Also documents Web access (query any website, with a caution that it can send workspace content to external services), Code Intelligence repository browsing, externally synced issues and comments, and that Loops run by default only on issues created within Linear unless specific external sources are enabled.

  Documents persistent actions available to a Loop including creating issues, updating projects and initiatives, editing Linear documents, adding summaries and follow-up work to a project document, posting Slack messages and starting a coding session, alongside example Loops that create and delegate work. The Coding sessions permission states that if enabled the Loop will be able to start a coding session to open a draft pull request. "Allow changes outside of triggering entity": when disabled the Loop can write data only on the entity that triggered the run; when enabled it can write to other entities within Team access scope; the setting does not apply to scheduled Loops.

  States Loops can find public Slack channels by name including Slack Connect channels and read their history, read messages from Slack links, post new messages, add or remove emoji reactions, and send one-way direct messages to full members; and that permitted MCP connectors can post comments or updates to external services such as Slack. Records Slack limits: no private channels, direct-message history or group direct messages; no workspace-wide message search, channel creation, message editing or deletion, file upload or Slack canvas management; replies in Slack Connect channels, DMs and unrelated threads do not continue a conversation; where automatic channel joining is disabled, only channels Linear has already joined can be used.

  Describes background execution after configuration and publication, with oversight through configured permissions and subsequent run history review; no per-run or per-action human approval or confirmation mechanism is documented anywhere on the page.

  Establishes layered permission mechanisms: Loop-scoped Linear team access and per-Loop permission toggles; Slack actions through the workspace's shared Slack connection rather than a personal connection; and MCP connectors that require the configuring user to authenticate and that use their personal account's connection. States a Loop can only use MCP connectors that are connected to the workspace, enabled for Linear Agent and allowed by workspace admins, with access also depending on the Loop's scope and permissions, and that an MCP connector can only access data and perform actions permitted by the MCP's configuration and the Loop's data scope.

  Documents workspace Loop management under Settings → Security → Workspace management → Manage loops, team Loop management under team Access and permissions, per-Loop "Who can edit" options of all members, workspace admins or team owners, or only the loop owner, and enable, disable and permanent delete actions. Documents MCP allowlisting by Owners (Enterprise) or workspace Admins (Business), the allowed external sources list configured by workspace owners, and the requirement that admins enable Code Intelligence, coding sessions and Slack Loop access before the per-Loop permissions can be used. Records limits: 100 Loops per team, 500 workspace-level Loops, 100 unpublished drafts per person, eight conditions, eight actions, 64-character names and 255-character descriptions.

  States Loops are available on paid plans, that a Loop must be explicitly created and published to exist, that draft edits apply only on Publish, and that permissions, integrations and connectors must be configured for the actions a Loop performs.

  States "Available on paid plans" and that Loops that use Linear Agent draw from the workspace's AI credits; applies no preview, beta or general-availability label.

  States Loops use AI credits purchased and managed by workspace admins, that purchased credits expire 12 months after purchase, that Loops pause if the workspace runs out of credits unless more are purchased or auto top-ups are enabled, and that Loops applying only configured issue updates can continue without AI credits.

  Documents run-failure causes (missing permissions, untrusted external sources, exhausted AI credits), the loop run failed notification, permanent deletion, and that removed connectors cannot be restored with a previous Loop version because they must be re-authenticated.

• Linear Agent

  Supports: Function · Data access · Default state · Admin controls · Primary source

  Describes Linear Agent as acting on workspace context to create and update issues, projects, milestones and initiatives, summarise and analyse work, answer questions and post, edit and delete its own comments — the agent that each Loop run invokes.

  States Linear Agent works with existing workspace data including teams and sub-teams, initiatives and projects with milestones and cycles, issues and their relationships, comments, activity history and documents.

  States Linear Agent is available by default in a workspace and that a workspace admin or owner can disable it in Settings → AI → Linear Agent; the page does not state what happens to already configured Loops after it is disabled.

  Documents the workspace-level Enable Linear Agent toggle available to admins and owners.

• Changelog: Introducing Loops

  Supports: Function · Availability · Admin controls · Licensing · Primary source

  Public introduction on 20 July 2026: Loops take on recurring work, are described in plain language and run on a schedule or in response to an event, using workspace and connected-tool context to decide on the next best action.

  States Loops were available on Business and Enterprise plans at launch and use AI credits.

  States Loops run at the team and workspace level with shared visibility and control, and that anyone with access can review their instructions, see how they are configured and inspect what happened during each run.

  Records the launch promotion of $20 per seat in promotional credits applied automatically to runs and expiring on 20 August 2026 — a temporary promotion, not the price of Loops.

• Changelog: Loops for product management

  Supports: Function · Actions · External actions · Licensing · Primary source

  Records the 14 September 2026 expansion, including that each loop run starts a conversation with Linear Agent and that once it finishes the conversation can be continued like any other agent session to review the work, add context or direct what happens next; and lists the new project, initiative, cycle and issue triggers.

  States that in addition to creating issues and updating projects and initiatives, Loops can now edit Linear documents, and can draft and post tailored Slack updates using context from Linear and connected tools.

  Documents drafting and posting tailored Slack updates as part of the same expansion.

  States introductory Loops credits were extended through 31 December, recorded as a temporary promotion.

• Slack integration

  Supports: External actions · Admin controls · Permission basis · Primary source

  Integration-side statement that Loops can connect to the workspace's Slack integration to work in public Slack channels: read message history from Slack links, post messages, send direct messages to full Slack members, add or remove emoji reactions, and continue conversations in public-channel threads started by a Loop.

  States that to connect Slack with a Loop an admin must enable Linear Agent and allow Loops access in Settings → Integrations → Slack → Workspace, after which Slack is selected from the Loop draft connector menu.

  Establishes that Slack Loop actions use the workspace's shared Slack connection rather than an individual's personal Slack authorization.

• AI Credits

  Supports: Licensing · Admin controls · Primary source

  States Linear's coding sessions and Loops features are billed on actual usage, that added funds expire 12 months after purchase, that promotional balances have their own expiration dates visible in the Usage and credits dashboard, and documents top-ups and Automatic Reload.

  Documents the workspace credit purchase, automatic reload threshold and usage dashboard controls available to admins.

• Coding sessions

  Supports: Actions · Limitations · Primary source

  Boundary: states that a coding session runs an agentic coding workflow that drafts a pull request and adds a diff to the issue for human review before merge — work performed by the coding session rather than by the Loop that started it.

  Supports the exclusion of downstream repository edits, commits and pull-request implementation from the Loop's own action boundary.

• MCP server

  Supports: General · Primary source

  Boundary: Linear's MCP server exposes Linear to external clients and is distinct from the MCP connectors a Loop may be permitted to call; this record covers only the latter.