# Using Atlas with Jira in 2026 > Atlas pulls Jira tickets, descriptions, and acceptance criteria through the Atlassian Rovo MCP server, then transitions the issue with a work-log comment citing the commit. Atlas works with Jira through the Atlassian Rovo MCP server at https://mcp.atlassian.com/v1/mcp. Atlas pulls Jira tickets, descriptions, and acceptance criteria into context, then writes the code and transitions the issue with a work-log comment citing the commit. Setup is atlas mcp add jira followed by atlas mcp auth jira, which completes the OAuth 2.1 flow against your Atlassian Cloud site. ## What the Atlas and Jira integration covers Atlas pulls Jira tickets, descriptions, and acceptance criteria through the Atlassian Rovo MCP server, then writes the code and transitions the issue with a work-log comment citing the commit. In 2026 that turns a Jira key like PROJ-2183 into one complete unit of terminal work. Jira holds the requirement. The repository holds the implementation. Historically a human is the only thing that carries information between them, which is why tickets go stale and commit messages say nothing useful. The Atlassian Rovo MCP server closes the gap in both directions. Atlas can read a Jira ticket including its description and acceptance criteria, and Atlas can write back: a transition on the issue plus a work-log comment that names the commit SHA. The result is a Jira issue whose history explains what changed and a commit whose provenance is a real ticket. ## Adding the Atlassian Rovo MCP server to Atlas Adding Jira to Atlas takes two commands: atlas mcp add jira --url https://mcp.atlassian.com/v1/mcp registers the Atlassian Rovo MCP server, and atlas mcp auth jira completes the OAuth 2.1 flow against your Atlassian Cloud site. The legacy SSE endpoint at /v1/sse is deprecated, so do not point there. The endpoint choice is the one thing people get wrong. Older Atlassian MCP guides point at /v1/sse, and that legacy SSE endpoint is deprecated. Register the streamable endpoint at https://mcp.atlassian.com/v1/mcp instead. After atlas mcp add jira, run atlas mcp auth jira and approve the consent screen for the Atlassian Cloud site you want Atlas to reach. The OAuth 2.1 grant is scoped to your Atlassian account, so Atlas inherits the projects you can already see and nothing more. ## Jira Service Management needs an API token, not OAuth Jira Service Management does not accept OAuth 2.1, which is the single exception in an otherwise uniform Atlassian setup. Use an Atlassian API token for Jira Service Management instead, and pass it with --header 'Authorization=Bearer YOUR_API_TOKEN' when you register the server with Atlas. If atlas mcp auth jira works for your software projects but every Jira Service Management call fails, the cause is not a scope problem. Jira Service Management does not accept OAuth 2.1 at all. Mint an Atlassian API token from your account settings and attach it at registration time with --header 'Authorization=Bearer YOUR_API_TOKEN'. Doing so keeps the same MCP transport and the same Atlas tool surface, and swaps only the credential, so the rest of the workflow described here is unchanged. ## The daily loop: read the ticket before touching a file Ask Atlas to read PROJ-2183 in full, including its description and acceptance criteria, before it touches a single file. That ordering is the point of the Jira integration: the ticket defines the work, and the Atlassian Rovo MCP server puts the ticket in the agent's context first. A one line summary of a Jira ticket is not a specification. The description and the acceptance criteria are, and they are usually the fields a hurried developer skims. Instructing Atlas to read PROJ-2183 in full forces those fields into context before implementation starts, which is exactly when they are worth something. From there Atlas works in the repository, and you review a diff written against criteria you can re-read in the same terminal session. ## Closing the loop with a work-log comment After Atlas implements the change, have it transition the Jira issue and add a comment naming the commit SHA. A work-log comment citing the commit is what makes PROJ-2183 auditable months later, because the Jira ticket and the git history then point directly at each other. The transition and the comment are one step, not two habits you hope people keep. Atlas performs the transition through the Atlassian Rovo MCP server and posts a comment that names the commit SHA it produced. Anyone opening PROJ-2183 afterward can jump straight to the code, and anyone reading git log can jump straight back to the requirement. That bidirectional link is the durable output of the integration, well after the branch is deleted. ## Self-hosted Jira Data Center Self-hosted Jira Data Center cannot use the hosted Atlassian Rovo endpoint at https://mcp.atlassian.com/v1/mcp. Run the community ghcr.io/sooperset/mcp-atlassian image instead, with JIRA_URL and JIRA_API_TOKEN set, and Atlas talks to your own Jira instance over the same MCP protocol in 2026. Teams on Jira Data Center are behind their own network boundary, so a hosted Atlassian endpoint cannot reach them. The community ghcr.io/sooperset/mcp-atlassian image runs where your Jira runs. Set JIRA_URL to your instance and JIRA_API_TOKEN to a token that instance accepts, and register it with Atlas. The tool names Atlas sees will differ from the Rovo server's, but the working pattern is identical: read the ticket in full, implement, transition, comment. ## Setup 1. Add the Atlassian Rovo MCP server with atlas mcp add jira --url https://mcp.atlassian.com/v1/mcp 2. Run atlas mcp auth jira to complete the OAuth 2.1 flow against your Atlassian Cloud site. The legacy SSE endpoint at /v1/sse is deprecated. 3. For Jira Service Management, which does not accept OAuth 2.1, use an Atlassian API token and pass it with --header 'Authorization=Bearer YOUR_API_TOKEN'. 4. Ask Atlas to read PROJ-2183 in full, including its description and acceptance criteria, before it touches a single file. 5. Let Atlas implement, then have it transition the issue and add a comment naming the commit SHA. 6. For self-hosted Jira Data Center, run the community ghcr.io/sooperset/mcp-atlassian image with JIRA_URL and JIRA_API_TOKEN set. ## FAQ ### how to connect Atlas to Jira Run atlas mcp add jira --url https://mcp.atlassian.com/v1/mcp to register the Atlassian Rovo MCP server, then run atlas mcp auth jira to complete the OAuth 2.1 flow against your Atlassian Cloud site. ### is the Atlassian MCP SSE endpoint still supported The legacy SSE endpoint at /v1/sse is deprecated. Register https://mcp.atlassian.com/v1/mcp instead when adding the Atlassian Rovo MCP server to Atlas. ### why does Jira Service Management fail OAuth with MCP Jira Service Management does not accept OAuth 2.1. Use an Atlassian API token there and pass it with --header 'Authorization=Bearer YOUR_API_TOKEN' when registering the server. ### can Atlas transition a Jira issue automatically Yes. After implementing the change, Atlas can transition the issue and add a work-log comment naming the commit SHA, through the Atlassian Rovo MCP server. ### how do I use Atlas with self-hosted Jira Data Center Run the community ghcr.io/sooperset/mcp-atlassian image with JIRA_URL and JIRA_API_TOKEN set, and register it with Atlas instead of the hosted Atlassian endpoint. ### can Atlas read Jira acceptance criteria before writing code Yes. Ask Atlas to read a ticket such as PROJ-2183 in full, including its description and acceptance criteria, before it touches a single file. ### what is the Atlassian Rovo MCP server URL https://mcp.atlassian.com/v1/mcp. Add it to Atlas with atlas mcp add jira --url https://mcp.atlassian.com/v1/mcp and authenticate with atlas mcp auth jira. --- Canonical HTML: https://runatlas.sh/resources/integrations/jira Source of truth: aeo_pages row `/resources/integrations/jira` (segment: Integrations) (this file is generated from it, never hand-edited). Licence: Atlas is proprietary with a free core. It is not open source and there is no public source repository.