# How to connect Linear to Claude: connector and MCP > Claude's Linear connector runs on Linear's own MCP server. Add Linear through Upsy to choose which actions Claude may take and see every change it makes. Updated 2026-10-17 · https://upsy.ai/guides/claude/linear ## Short answer Claude has a Linear connector, made by Linear. Open **Customize › Connectors**, add Linear from the directory and sign in. It takes a minute, on every plan, Free included. It runs on Linear's own MCP server, the same one Upsy uses, so for many people it's enough. Use Upsy instead if you want limits that hold in Claude Code too, a record of every change, or one Linear connection for all your assistants. ## The built-in way ### Claude's Linear connector The [directory page](https://claude.com/connectors/linear) lists Linear as made by Linear Orbit, Inc., with the address `https://mcp.linear.app/mcp`. That's [Linear's MCP server](https://linear.app/docs/mcp), a connector that lets an assistant find, create and update issues, projects and comments. The page lists 22 tools, but the server offered 76 when we connected it in October 2026, including documents, initiatives and your Linear inbox. Web connectors work for all users in Claude, Cowork, Claude Desktop and the mobile apps ([Anthropic's help](https://support.claude.com/en/articles/11176164-use-connectors-to-extend-claude-s-capabilities)), and Claude reaches only what your Linear account can open. On Team and Enterprise, an Owner first adds Linear under **Organization settings › Connectors**, clicks **Browse connectors**, then **Add to your team**. Each member then signs in to Linear with their own account. ### Claude Code Claude Code picks up your claude.ai connectors, so Linear shows up there as "claude.ai Linear". You can also add Linear's server yourself with `claude mcp add --transport http linear-server https://mcp.linear.app/mcp`, then run `/mcp` to sign in ([Linear's docs](https://linear.app/docs/mcp)). ### Where it stops - **Per-tool limits, inside Claude only.** Under **Customize › Connectors**, Linear's tools are grouped into read-only and write/delete, and each can be **Always allow**, **Needs approval** or **Blocked**. On Team and Enterprise, an Owner can set them for everyone. Anthropic's own example: Claude may view Linear issues but not create them or change their status. - **Those settings skip Claude Code.** A [June 2026 report](https://github.com/anthropics/claude-code/issues/64521) found that Claude Code in the terminal ignores the approval settings from claude.ai and uses its own allow, ask and deny rules. - **A read-only address exists, but it's all or nothing.** Linear also runs `https://mcp.linear.app/mcp/readonly`, which only offers read tools. You can add it as a custom connector instead of the directory one. On Free, that uses your one custom connector. - **One sign-in.** The connector holds one Linear sign-in. Linear says each workspace needs its own sign-in. - **No record outside the chat.** What Claude did in Linear lives in each conversation, with no list across chats. ### Where Upsy fits - **Limits Upsy enforces itself.** Read-only, or only the Linear tools you pick. Upsy refuses the rest on its side, so the limit holds in claude.ai, Claude Desktop and Claude Code alike. - **History:** every Linear tool Claude runs is listed in Upsy, with the issue or search it was about. - **More than one Linear account:** up to five, each with its own access. - **One setup:** connect Linear once, and Claude, Claude Code, ChatGPT, Grok and Muse all use it, each at its own level. - **Your other apps in the same chat,** such as turning a Gmail thread into an issue. ## What you need - **A Linear account.** Linear's docs list no plan requirement for its MCP server. - **Claude on any plan.** Upsy is a custom connector, and Free allows one. - **On Team or Enterprise, an Owner.** They add Upsy once under **Organization settings › Connectors** (**Add › Custom › Web**), then you click **Connect**. - **Upsy:** a free account. ## Connect Linear to Claude 1. Sign up at [upsy.ai](https://upsy.ai) with **Set up Upsy**. 2. In Upsy, open **Apps**, click **Connect an app** and pick **Linear**. Linear's consent screen opens. Approve access for the workspace you want. There's no API key to create. 3. Open **Assistants**, click **Add assistant** and pick **Claude** under **Which assistant?**. Copy your Upsy address: ``` https://upsy.ai/connect ``` 4. In Claude, go to [Customize › Connectors](https://claude.ai/customize/connectors). Click **Add**, hover over **Custom** and choose **Web**. Some accounts show an **Add custom connector** button instead. 5. Name it Upsy, paste the address and click **Continue** twice, keeping the sign-in settings Claude detects. Then click **Add**. 6. Click **Connect** next to Upsy. Sign in on the Upsy page that opens. Linear is ticked and set to All; switch it to **Read-only** if Claude should only look. Click **Allow**. 7. In a chat, click **+**, hover over **Connectors** and check that Upsy is on. Each screen is in [how to add a custom connector to Claude](/guides/claude/custom-connector). To allow only some Linear actions, use **Pick actions** on Claude's page under **Assistants**. Already using Upsy in Claude? Just connect Linear on the **Apps** page. It starts at Full access for the assistants you've added, so start a new chat, and lower the level on Claude's page if you want. ## Check that it works ``` Using Upsy, which Linear workspace am I connected to? Then list my open issues, highest priority first. ``` Claude should name your workspace, then list issues with identifiers such as ENG-142. It may ask before running Upsy's `app_read` tool the first time. In Upsy's **History**, you'll see "linear get workspace" and "linear list issues" under Claude. ## What Claude can do with Linear Upsy passes on Linear's own tools. Claude lists them with `app_tools`, runs the reading ones with `app_read` and the rest with `app_call`. **With Read-only**, Claude can: - Find issues by team, status, assignee, label, cycle, project or text, and open one in full, with its attachments and git branch name. - Read projects, milestones, initiatives, cycles, documents and comments. - Look up teams, people, statuses, labels, templates and saved views. - Read project and initiative updates, your Linear inbox, and releases where your workspace uses them. - Search Linear's own help docs. **With Full access**, it can also: - Create and update issues: title, description, status, assignee, priority, labels, due date, estimate, project, cycle and parent. - Link issues as blocking, blocked by, related or duplicate. - Comment on issues, projects and documents. - Create and update projects, milestones, initiatives, documents and status updates. - Share an issue with a teammate, and mark inbox items read or snooze them. - Delete comments, attachments and status updates. When we checked, no tool deleted an issue or a project. - Review and merge pull requests in Linear's code review, which Linear calls diffs. **With Pick actions**, you tick tools one by one. Note that Linear uses one tool, `save_issue`, to both create and update issues, so ticking it allows both. Leave out `merge_diff` unless Claude should merge code. Any tool Linear adds later stays off until you tick it. Read-only allows only tools Linear marks read-only, and never one whose name says it changes something, such as save or delete. ## Prompts to try - "Write my stand-up from Linear: what I moved to Done since yesterday, and what's In Progress now." - "Summarize the last finished cycle for the Web team. What shipped, what slipped, and which bugs we closed?" - "Read this spec and propose a Linear project with milestones and issues. Show me the list before you create anything." - "Comment on ENG-142 with a short summary of what we decided in this chat." ## Troubleshooting ### "Your account was authorized, but no MCP server was found at the provided URL" Claude shows this right after the Linear sign-in. In [a June 2026 report](https://github.com/anthropics/claude-ai-mcp/issues/484), Anthropic traced it to a connector saved with Linear's old `/sse` address, which Linear has removed. In Claude Code, the same cause shows as ["DEPRECATED TRANSPORT"](https://github.com/anthropics/claude-code/issues/61429) on every Linear call, and reconnecting doesn't fix it. Remove the Linear connector in **Customize › Connectors** and add it again from the directory, which uses `/mcp`. ### "MCP server 'linear' skipped — same command/URL as already-configured 'claude.ai Linear'" You installed Linear in Claude Code while the claude.ai connector also exists. [It's informational](https://github.com/anthropics/claude-code/issues/39511): both reach the same server. Disconnect one if the message bothers you. ### Linear says "not connected" after /compact in Claude Code A [September 2026 report](https://github.com/anthropics/claude-code/issues/95351), still open, describes this in the desktop app. Disconnecting and reconnecting Linear under Connectors brings the tools back in the same session. ### "The user hasn't allowed you to use Linear's "save_issue"" That's Upsy refusing a tool outside Claude's level, such as a create or update on Read-only. Change Linear's level on Claude's page in Upsy if Claude should be able to do it. ### "Linear access expired. Ask the user to reconnect it in Upsy" On Upsy's **Apps** page, click **Reconnect** next to Linear and approve again. ### Claude uses its own Linear connector instead of Upsy With both connected, Claude may pick either. Say "using Upsy", or switch the Linear connector off for that chat under **+ › Connectors**. ## Security and privacy - **Linear's own sign-in.** Upsy connects to Linear's MCP server with your Linear login, and the token stays in Upsy. Claude calls Upsy, Upsy calls Linear, and the level you set narrows what Claude may do. - **History:** every Linear tool Claude runs is listed under Claude, and so are refused attempts. No one can edit or delete it. - **Alerts:** an email when one assistant opens 100 or more items in 10 minutes or deletes 50 or more, and when a new assistant or app is connected. - **Claude's prompts are the check.** Upsy has no approval step of its own. It marks `app_read` as read-only and `app_call` as able to change things, so Claude's tool permissions can treat them differently. - **Turning it off:** set Linear to No access for Claude, click **Remove Claude** on its page in Upsy, or disconnect Linear on the **Apps** page. ## FAQ ### Does Claude have a Linear integration? Yes. Linear is in Claude's Connectors Directory, built by Linear, on every plan. Upsy adds per-action access and a history on top. ### Is Claude's Linear connector an MCP server? Yes. It connects to `https://mcp.linear.app/mcp`, Linear's hosted MCP server. Upsy connects to the same server. ### Can Claude create Linear issues? Yes, with either one. Through Upsy, Claude needs Full access, or Pick actions with `save_issue` ticked. ### Can I make Claude read-only in Linear? Yes. Set Linear's write/delete tools to Blocked under **Customize › Connectors**, or add Linear's `/mcp/readonly` address as a custom connector. Through Upsy, set Linear to Read-only for Claude; Upsy refuses changes itself, in Claude Code too. ### Can Claude use two Linear workspaces? With Upsy, yes, up to five Linear accounts with different email addresses. Upsy tells Linear accounts apart by email, so signing in again with the same email replaces the earlier connection. ### Does it work in Claude Code? Yes. Run `claude mcp add --transport http upsy https://upsy.ai/connect`, then `/mcp` to sign in. Claude Code gets its own level in Upsy. See [Claude Code MCP servers](/guides/claude-code). --- Upsy connects personal AI assistants (Muse, Grok, ChatGPT, Claude, Cursor, Claude Code, Codex, Devin) to your Gmail, Google Calendar, Google Drive, Slack, Notion and other apps through one MCP server, with a shared memory, files and a cloud browser signed in to your sites. You choose what each assistant may do in each app, see everything it did, and turn it off in one click. Upsy is free. To add Upsy to an assistant, add a remote MCP server (Streamable HTTP) named Upsy at https://upsy.ai/connect. It signs in with OAuth; there is no API key.