Atlassian MCP
Atlassian Rovo MCP connects agents to Jira, Jira Service Management, Confluence, Bitbucket, Compass, and Teamwork Graph workflows while respecting existing user permissions.
Verdict
Common use cases
- Triage support tickets from chat
- Draft sprint retrospectives from Jira data
- Search Confluence docs during planning
- Update issue status without switching tabs
- Pull burndown metrics for standups
Integration
- Vendor
- Atlassian MCP
- Category
- other
- Auth
- OAUTH2
- Composio slug
atlassian_mcp
Tools
Setup
Setup guide
- 11. In Switchy, open your workspace settings and navigate to the Integrations tab. 2. Find Atlassian MCP in the directory and click Connect. 3. You'll redirect to Atlassian's OAuth consent screen — sign in with an account that has access to the Jira sites and Confluence spaces your team needs. 4. Grant the requested scopes (typically read/write for Jira issues, read for Confluence pages, and user profile access). 5. Atlassian redirects you back to Switchy with a success confirmation. 6. Open any Project, type '@Atlassian' in the chat, and ask it to list your open Jira issues — if it returns results, the connection works. 7. If you see an authorization error, check that your Atlassian admin hasn't restricted third-party app access in your organization's security settings.
What teammates see: by default, memories from Atlassian MCP are scoped to the Project (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Projects have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
Sprint Summary
@Atlassian pull all issues closed in the current sprint and summarize what the team shipped, grouping by epic or component.
Ticket Triage
@Atlassian list all unassigned issues in the support queue, then rank them by priority and explain your reasoning.
Confluence Search
@Atlassian search Confluence for pages about our authentication flow and summarize the key implementation details.
Blocker Report
@Atlassian find all blocked issues in the engineering board and write a brief report explaining each blocker and its impact.
Update Issue Status
@Atlassian transition issue PROJ-123 to 'In Review' and add a comment that the PR is ready for feedback.
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@atlassian show me all open issues assigned to me across projects, grouped by priority
Here's an illustrative summary of your open issues: **High Priority (3):** PROJ-142 (API rate limit handling), TEAM-89 (Database migration blocker), INFRA-201 (Production deployment failure). **Medium Priority (7):** Issues span feature requests and minor bugs across three projects. **Low Priority (2):** Documentation updates. This example assumes the MCP fetched from Jira; actual output depends on your Jira instance configuration and project visibility.
This read-heavy example demonstrates cross-project aggregation. Requires OAuth scope for Jira read access. The MCP must be authorized for all projects you want visible — private projects or restricted boards won't appear unless explicitly granted during OAuth setup.
@atlassian create a new Confluence page in the Engineering space titled 'Q1 Incident Retrospective' with a basic template
Page created successfully at https://yourinstance.atlassian.net/wiki/spaces/ENG/pages/123456789. The page includes a standard retrospective template with sections for timeline, root cause, and action items. This illustrative example shows a write operation; the actual page ID and URL depend on your Confluence space structure and permissions.
This write example shows content creation capability. Requires OAuth scope for Confluence write access to the target space. Be aware that page creation is immediate and irreversible through the MCP — you'll need to manually archive or delete pages created in error. Confluence space permissions apply.
@atlassian review the last 15 comments on PROJ-142 and draft a summary for stakeholders explaining the current blocker
Based on the comment thread (illustrative): The blocker centers on third-party API rate limits hitting 429 errors during peak hours. Engineering proposed two solutions: implement exponential backoff (2-day effort) or negotiate higher rate limits with the vendor (timeline uncertain). Product team favors the backoff approach to unblock the release. Next step: finalize implementation plan in tomorrow's standup. This example shows the AI synthesizing Jira comment data fetched by the MCP.
This synthesis example pairs the MCP's data retrieval with the AI's reasoning to distill technical threads into stakeholder-friendly summaries. Requires read access to the issue and all comments. The AI interprets comment content — accuracy depends on comment clarity and completeness in the source issue.
Use-case deep-dives
When Atlassian MCP fits distributed sprint retros
A 6-person engineering team runs biweekly retros and wants to pull velocity trends, blocker patterns, and cycle time from Jira without opening five browser tabs. The Atlassian MCP handles OAuth2 once, then lets the team query sprint metrics, link types, and issue histories in a shared Switchy thread. This works if your Jira instance has under 20 active projects and you're asking for aggregated data, not real-time updates. The MCP shines when the retro lead needs to compare the last four sprints or surface recurring labels—queries that take 10 clicks in the Jira UI. If your team runs retros async and references the same datasets weekly, the Atlassian MCP saves 15 minutes per session and keeps the conversation in one place.
Atlassian MCP for support-to-eng handoffs
A 3-person support team escalates bugs to engineering through Jira Service Management, but the context always gets lost in translation. The Atlassian MCP lets support reps draft the escalation ticket in Switchy, pull the customer's last 10 interactions from JSM, attach the relevant Confluence runbook, and hand the whole thread to eng with one OAuth login. This scenario works if your support queue stays under 200 tickets per week and your runbooks live in Confluence. The MCP breaks down if you need real-time SLA tracking or multi-project dependencies—it's built for handoff clarity, not workflow automation. If your escalations currently require three Slack messages and two Loom videos, the Atlassian MCP collapses that into a single shared context that both teams can reference.
When Atlassian MCP speeds up new-hire ramp
A 10-person startup onboards two engineers per quarter and maintains 40 Confluence pages covering architecture, deploy processes, and team norms. New hires spend their first week asking the same questions in Slack because search in Confluence returns 200 results. The Atlassian MCP connects to Confluence via OAuth2 and lets the onboarding buddy pull the exact 3-4 pages that answer the new hire's question in a Switchy thread, then annotate them inline. This works if your Confluence space has under 500 pages and clear page hierarchies. The MCP struggles if your docs are scattered across 10 spaces or buried in comments. If your new hires currently take 5 days to find the deploy checklist, the Atlassian MCP cuts that to 20 minutes and creates a reusable onboarding thread for the next hire.
Frequently asked
What does the Atlassian MCP do in Switchy?
It connects your Switchy workspace to Atlassian products like Jira and Confluence, letting AI assistants read project data, search documentation, and reference work items during conversations. Since the tool list isn't finalized, expect basic read operations on issues, pages, and project metadata. You won't manage sprints or edit content directly through the MCP.
Do I need admin access to connect Atlassian via OAuth?
You need permission to authorize OAuth apps in your Atlassian organization. Most Atlassian Cloud instances let standard users approve read-only integrations, but write scopes typically require admin approval. Check with your Atlassian admin before connecting if your organization enforces strict app policies.
Can the Atlassian MCP create or update Jira issues?
Not yet confirmed — the current tool set is incomplete. Atlassian's API supports writes, but most MCP integrations start with read-only operations. If you need to create issues or update tickets, use Jira's native interface or a dedicated automation tool until Switchy documents write capabilities for this MCP.
How does this compare to querying Atlassian APIs directly?
The MCP abstracts away pagination, authentication refresh, and response parsing. You get natural-language queries instead of constructing JQL or REST calls. Trade-off: you lose fine-grained control over field selection and can't access every endpoint Atlassian exposes. Use the MCP for conversational access, APIs for custom workflows.
Who on the team should connect the Atlassian MCP?
Whoever owns your team's Atlassian account and understands which projects the AI should access. That person's OAuth token determines visibility — the MCP inherits their permissions. If your Jira has confidential projects, connect with a service account or a user whose access matches your team's needs.