Turn Linear into a controlled workspace for every agent
Planning agents need context; delivery agents sometimes need to change the plan. AIP sits in front of Linear's official MCP server so each agent sees the exact mix of issue, project, comment, and document tools its role needs.
What your agent can do
Read tools locate teams, cycles, issues, projects, comments, and documents. The save tools create or update shared Linear objects, so AIP lets you place them on Confirm or hide them while keeping the research surface available.
| Tool | What it does | Access |
|---|---|---|
| list_issues | Search and filter issues across accessible Linear teams and projects. | read |
| get_issue | Retrieve one issue with its description, state, relations, attachments, and documents. | read |
| save_issue | Create an issue or update fields, relationships, labels, assignment, and state. | write |
| list_projects | List projects with filters for team, state, dates, lead, and other project metadata. | read |
| get_project | Read one project and its current planning and status information. | read |
| save_project | Create or update a project, including teams, dates, labels, links, and description. | write |
| list_comments | Read comments attached to an issue, document, project, initiative, or status update. | read |
| save_comment | Create or update a comment on a supported Linear object. | write |
| list_documents | Search and list Linear documents and their owning teams, projects, initiatives, or issues. | read |
| get_document | Fetch a document by ID or slug with its complete Markdown content. | read |
| save_document | Create or update a Linear document and attach it to the correct owner. | write |
| list_cycles | List team cycles and their dates, progress, and completion state. | read |
| list_teams | Resolve the teams available to the connected Linear user. | read |
Linear's hosted MCP server evolves and can expose a larger workspace-aware inventory. These are representative exact tools verified through AIP's current connected discovery, not a frozen promise of the full server.
Capability snapshot: 1 September 2026 · provider source
Let research travel farther than mutation rights
A triage agent can read issues, projects, cycles, comments, and documents to build an accurate picture without changing the workspace. A delivery agent may also create issues or update a status. A planning agent might write project documents but never touch labels or assignments. Those are different roles, even when they share the same Linear account.
Model them directly in the AIP control room. The official Linear server remains responsible for workspace data and OAuth; AIP controls which of its discovered tools reach each agent and account. The integrations index shows the same pattern for adjacent systems such as Notion, Sentry, and PostHog.
Connect Linear once through OAuth
- Choose Linear in the AIP dashboard. AIP opens Linear’s hosted OAuth flow; approve the workspace identity you want to use.
- Add that account to an agent and inspect the discovered tool list. Allow the reads its role needs, set relevant saves to Confirm, and hide unrelated workspace administration.
- Paste the generated agent URL into any MCP client. Claude, Claude Code, Codex, Cursor, and other clients all reach the same scoped Linear connection instead of keeping their own provider credentials.
Linear’s server is Streamable HTTP, while AIP can also publish the scoped agent as OpenAPI. The developer overview explains the two client surfaces; neither changes the underlying Linear permissions.
Use issue context before creating issue noise
Reliable automation reads the parent, relations, documents, and current
comments before it writes. That prevents duplicate issues and keeps a plan in
its existing document rather than scattering revisions across comments. When
you grant save_issue, include those checks in the agent’s instructions and
keep the final mutation reviewable.
For multiple workspaces or separate planning and delivery agents, review the pricing options rather than sharing one broad agent. Independent account labels and tool policies make both the prompt context and later audit trail easier to understand.
What to ask your agent
Linear prompts work best when they name a team or project, a time boundary, and whether the result should be analysis or a mutation: