Connect Linear

Connect Linear so the assistant can look up issues in your workspace when a customer asks about one. An organization administrator connects the workspace.

• Connects: a Linear workspace.
• Tools from: Linear's MCP server, read when you connect.
• Sign-in: Linear, in your browser.
• Replies: replies in a conversation that used its data wait for your approval unless you turn on Answer without approval. See how replies are approved.

What the assistant doesCopy Anchor Link

• Looks up issues in the connected workspace with the Linear tools you turn on.
• Answers from what it finds, and describes a change only when Linear confirms it.

Connect your workspaceCopy Anchor Link

Open Integrations in your organization's dashboard sidebar and locate the Linear card.

1. Choose Connect on the Linear card.
2. Continue to Linear and approve access to the workspace you want to connect.
3. Return to Samva. The Linear integration page confirms the connection and lists the account. Choose Add workspace to connect another one.

To let an agent use it, open the agent's Tools section and switch on Let this agent use it for Linear. Through the API, add a grant for the integration to the agent's integrations and publish. See What an agent may use from an integration.

ToolsCopy Anchor Link

Linear's own tools are read from its MCP server when you connect a workspace, and kept current while the integration is installed. They appear under Agent tools on the Linear integration page after you connect. Each starts off: turn on the tools agents may use, then grant them in each agent's Tools section. A tool Linear adds later also starts off.

A tool Linear marks read-only reads, and it runs. Any other tool changes something or can't be undone, and it asks a person first until an organization owner lets it run without approval. See approving tool calls.

Approving tool callsCopy Anchor Link

A tool that only reads runs. A tool that changes something or can't be undone waits for a person's approval in Needs you before it runs, for every agent, whatever its preset. This holds on every Linear account.

An organization owner can change that for one tool. On the Linear integration page, each enabled tool shows Asks first or Runs without approval; switch it per tool, and the change applies to every agent that uses Linear. Let a tool that changes something run without approval, or make a tool that only reads ask first when it does more than its description says. Choosing Always allow when you approve a held call lets that tool run, for owners only. Through the API, send PATCH /v1/integrations/{integrationId} with { "tool": { "name": "<tool name>", "runsWithoutApproval": true } } to let a tool run, or false to make it ask first.

If Linear later changes whether a tool reads, changes something, or can't be undone, the owner's choice for it ends and the tool follows its new default: a tool that changes something asks first, a tool that only reads runs. A change that keeps that, such as new arguments, keeps the owner's choice. A tool Linear adds later starts at its default. To make one agent keep asking for a tool an owner let run, set that tool to Ask a person (or Never) in the agent's Tools section; that holds for that agent only.

Replies that use Linear dataCopy Anchor Link

By default, once the assistant has read results from Linear in a conversation, each reply it drafts in that conversation waits for your approval in Needs you, which shows Linear as the reason. A Linear call that failed, or that only proposed an action for a person to approve, adds no hold. This applies on every channel the agent answers on, and Always allow does not lift it. After that conversation reads Linear results, the agent also saves no new customer facts from it to shared memory, so another conversation never answers from them without this hold.

To stop holding replies for Linear data, an organization owner turns on Answer without approval on the Linear integration page. Through the API, send PATCH /v1/integrations/{integrationId} with { "answerWithoutApproval": true }. Those replies then follow your agent's reply rules like any other reply. Letting a tool run without approval does not change this.

ScenariosCopy Anchor Link

• When a customer asks, "Is the export bug fixed yet?", the assistant looks up the issue that tracks it and answers from its status.
• When a customer asks about a feature they requested, the assistant finds the issue that tracks it and tells them where it stands.

LimitsCopy Anchor Link

• Read only. Samva asks Linear for read access, so the assistant cannot create, update, or delete issues.
• An agent uses only the Linear tools you turned on and granted it, in a workspace you connected.
• Issue content is information for the assistant, never instructions to it.

Disconnect or reconnectCopy Anchor Link

Choose Disconnect on the Linear integration page to remove the account and its linked integrations. Agents that used it lose access; see when an integration is removed. If the connection needs attention, choose Reconnect to authorize it again.