Connect Linear, Slack and Notion to Cursor: Setup Steps

Two mechanisms, and knowing which one you are in saves a lot of clicking. Linear, Slack and Notion are Cloud Agent integrations, so connecting them also means completing Cloud Agent setup. Everything else works through MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition..

That split reads like a coverage list, as though MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. were the fallback for tools Cursor has not got to yet. Linear, Slack and Notion all turn up as MCP servers people connect as well, which the MCP guide goes through. What separates the two paths is which side starts the work. A Cloud Agent integration lets the issue or the thread launch an agent, while MCP hands a tool to an agent you started yourself.

Cloud Agent setup is where that difference turns into work. An agent triggered from a Linear comment needs somewhere to run, on a machine with your repository on it, and step four is what stands that machine up. An MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. server rides along in a session you already have open, which is why its setup asks for a command or a URL and whatever keys the server needs, and nothing about billing.

Confirm which of the two paths your tool is on, then decide which conversations you are willing to let start agents. Notion adds a third check, since its plan tier and the GitHub connection both have to be right before you create anything.

Notion's prerequisites are probably the place to start, and the reason is where they sit in the documentation. Cursor lists them after the four numbered steps: a Notion Business or Enterprise plan, and GitHub connected to your Cursor account. Read the page top to bottom and you create a Cursor User API Key, paste it into Notion, and only then find out the workspace is on a plan that cannot run the agent. The plan requirement is the item most likely to stop you, and it is the one you can check without touching Cursor at all.

Which tool to connect first depends on how many people are in the room. My first instinct is Slack, because work tends to get described there before it gets filed anywhere. That is too loose as a rule: it holds while you can still name every channel the mention might fire in. Past that, Linear is the easier first connection, since step three makes you pick a team and the four steps on Cursor's help page give Slack no equivalent narrowing step.

Step four reads identically in the Linear and Slack flows: connect a repository provider, enable usage-based pricing, confirm privacy settings. What Cursor's page does not say is whether the second connection re-runs it, so budget the time once and treat a repeat as possible.

Usage is billed the same way whichever one you connect. Linear and Slack both route you through the usage-based pricing step, and Notion usage is billed like Cloud Agent usage. On a shared channel the people starting agents will not always be the people watching the number, and that gap, more than the per-run cost, is what spend alerts are for.

Alerts email you when a threshold is crossed and do not stop anything. Spend limits are the separate control that caps spend.

Four steps from the integrations dashboard. The team you select in step three decides which Linear team the agent can be reached from.

1. 1Go to Cursor integrations at cursor.com/dashboard/integrations.
2. 2Click Connect next to Linear.
3. 3Authorize your Linear workspace and select a team.
4. 4Complete Cloud Agent setup: connect a repository provider, enable usage-based pricing, confirm privacy settings.

Once connected there are two intake routes, and they suit different moments. Delegating a Linear issue hands the agent something already written down. Mentioning @Cursor in a comment starts a Cloud Agent from the thread where the work is still being discussed.

Whether you can add teams afterwards is not something Cursor's help page settles, and I have not tested it. If your Linear workspace runs a team per squad, read step three as choosing whose comments can start agents.

The same shape as Linear, with a Slack app install in place of the workspace authorisation.

1. 1Go to Cursor integrations at cursor.com/dashboard/integrations.
2. 2Click Connect next to Slack.
3. 3Install the Cursor app in your Slack workspace.
4. 4Complete Cloud Agent setup: connect a repository provider, enable usage-based pricing, confirm privacy settings.

After that, mention @Cursor in any channel with a prompt to start a Cloud Agent. Inside a thread that already has an agent, asking in natural language works (@Cursor start a new agent to fix the login bug), and so does @Cursor agent [prompt]. @Cursor help lists the full command set.

The four steps do not spell out what changes for everyone else. Once the app is in, the trigger is a mention in any channel, so any channel is an intake surface whether or not you told people about it. Which is fine on a small workspace, where that is the point. In a large one, say which channels you expect it in when you announce it.

Notion is the exception: it is in beta and you set it up from Notion rather than from the Cursor dashboard. You will need a Cursor User API Key, which you create in the Cursor dashboard first.

1. 1In Notion, go to Agents in the sidebar, then New Agent, then Cursor.
2. 2Pick a starting point: a template or from scratch.
3. 3Create a Cursor User API Key at cursor.com/dashboard/api and enter it into Notion.
4. 4Share the pages and databases the agent needs.

The key comes from cursor.com/dashboard/api, a different page from the integrations dashboard the other two use.

The prerequisites bite before step one. You need a Notion Business or Enterprise plan and GitHub connected to your Cursor account. The integration itself is available to all Cursor users at no extra cost, and usage is billed like Cloud Agent usage. Once it is set up, @mention the agent in comments or assign tasks to Cursor from a task board.

Sharing is the access decision, and it is easy to read as housekeeping. What you share is what the agent can see, so start with the single page or board you want worked on and widen it once that has gone well.

Most third-party integrations work through MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition.. Browse the available servers in Customize in the sidebar, or on the Marketplace. If nothing exists for your tool, building a custom MCP server is the supported path rather than a workaround. For the three connectors above, the behaviour reference sits apart from the setup steps: Linear, Slack and Notion.

Since August 3, 2026 the Marketplace route also covers Google Workspace. Cursor publishes its own Gmail, Google Drive and Google Calendar plugins, each listed as Verified by Cursor with a one-line install (the Gmail page gives /add-plugin gmail) and an MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. definition you can read in the cursor/plugins GitHub repo before you connect the account. They are plugins rather than the connectors above, so they do not start cloud agents from a mailbox; they give a running agent the ability to search, read, draft and manage mail, files and events. The integrations guide has the action list per pluginA Cursor marketplace package that bundles MCP servers and skills (sometimes sub-agents and hooks); one click installs all of it into your Cursor instance. Press Enter for the full definition. and a safe first use for each.

Cursor's documented answer for a tool nobody has covered is to build the server yourself. That is a real project. A custom MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. server is software your team then maintains, and the MCP guide covers where its config lives and what tends to break. Read through the Marketplace listings and the community servers before you commit to writing one.

Split it in two. The intake half is the mention or the delegated issue; the execution half is the Cloud Agent that runs afterwards. They fail for different reasons, and the fixes are in different places.

In Slack the cheap probe is @Cursor help, which is documented to return the full command list. A reply is good evidence the app is installed and reachable, though it says nothing about whether the Cloud Agent side is healthy. That moves the question over to the Cursor side, where the cloud agent setup guide is the place to work through the environment. If nothing comes back at all, look at the app install before anything else.

Notion narrows faster than the other two. Check two things, neither of which is really a bug: a prerequisite is missing, or the page was never shared with the agent.

In Linear, rule out the team selection first. Checking which team you authorised is cheaper than going through an environment.