Atlassian AI agent: Rovo, Jira support, and eesel CLI

Which Atlassian AI agent are you evaluating?

This article is published by eesel, which offers a support teammate that integrates with Jira Service Management. I would compare the actual workflow in each product before deciding which one fits.

Rovo: search, chat, and agents

Rovo brings together search, chat, and agents across Atlassian and connected sources. Agents can be selected for specific tasks, and Rovo Studio provides a way to build and configure them. Atlassian's Rovo walkthrough explains these entry points.

For example, a team looking for project context is evaluating a different job from a support team deciding who can send customer replies. Both may use AI, but the required sources and allowed actions differ.

Do not treat Rovo as a future feature or only a search box. Equally, do not assume that selecting an agent grants every operation you need. Inspect its configuration and test the intended task.

Jira Service Management's virtual service agent

The virtual service agent is a support-focused option. Atlassian's setup documentation requires Service Collection Premium or Enterprise, AI activation by an organization admin, and a space admin.

It is also distinct from displaying a support-focused Rovo agent in a portal. The same guide says portal Rovo agents are currently available for Atlassian Accounts, whereas the virtual service agent supports all account types with portal access. Check the audience you actually serve before choosing between them.

The buying question is therefore specific: which support experience reaches your customers, on your subscription, with the right access rules?

Can Atlassian AI use knowledge outside Confluence?

Yes. Rovo connectors support third-party apps such as Google Drive, SharePoint, and Teams. That makes “all your documentation must move into Confluence” an inaccurate reason to choose another product.

The setup still needs attention. For Google Drive, an administrator configures the connector, and team members connect their accounts so results respect their access.

I would test three questions with the actual users who will rely on the system:

• Can they retrieve the current approved policy?
• Are they prevented from seeing restricted material?
• Can they tell which source supports the answer?

A connector's presence answers none of those questions by itself. Compare retrieval quality and permissions with representative content, not just the number of supported apps.

Pricing: distinguish the subscription from AI usage

There is no useful single price for “the Atlassian AI agent.” The relevant bill depends on the offering and feature.

Atlassian's Rovo credit guide lists these included credits per user per month:

Checked 8 September 2026. These are credit allowances, not subscription prices or a count of successfully resolved support tickets. Do not apply the standalone Jira allowance to Service Collection.

For a support rollout, separately confirm the virtual service agent's assisted-conversation terms and any additional usage charges. Use your current quote and administration screens for the subscription cost; a fixed per-agent number copied from an older article may not match your team size or billing term.

A sensible comparison includes license cost, usage, administration, and the human time needed to review or finish work. The support cost guide explains why the software line alone is insufficient.

Where eesel fits alongside Jira

eesel provides a support teammate that can read issue data and perform supported actions through its Jira Service Management integration. The CLI lets you operate that teammate from your existing terminal workflow.

The distinction is not “Atlassian-only knowledge versus external knowledge.” Both options can use external sources. For eesel, the relevant questions are whether its support workflow fits your queue and whether CLI access makes configuration, testing, and review easier for your team.

The JSM integration reference describes these capabilities:

JSM does not sync old issues into eesel's knowledge ahead of time. Live lookups are different from a searchable historical knowledge import. Connect policy sources such as Confluence, or upload approved files, alongside the Jira connection.

An assignment also needs an Atlassian account ID, not just a display name. These details matter more to a pilot than a broad claim that an agent “works with Jira.”

Inspect and test Jira support through eesel CLI

The CLI operates the same eesel agent and workspace as the dashboard. A support manager can review work in the dashboard while an engineer starts the evaluation from a terminal. Scripts and coding agents can also read the CLI's JSON output.

This is an interface to your configured eesel teammate, not a separate local agent or a generic command wrapper for every Jira endpoint.

Confirm your workspace and existing configuration

With Node.js 18.17 or newer, use the documented npm package. For an existing workspace:

npx @eesel/cli login
npx @eesel/cli whoami
npx @eesel/cli agents
npx @eesel/cli integrations --agent AGENT_ID
npx @eesel/cli instructions --agent AGENT_ID
npx @eesel/cli automations --agent AGENT_ID

Replace AGENT_ID with the intended agent ID. Check the workspace before changing anything. Existing automations matter: an agent already connected to a live queue may have work enabled.

If Jira is not connected, use eesel's integration setup in the dashboard. The documented form asks for the site subdomain, the Atlassian account email, and an API token. Enter the token in the setup form, not a shell command that may be saved to history. This connection is not an Atlassian Marketplace installation or an OAuth flow.

Enabling a Jira trigger makes eesel create a “Send web request” rule in Jira Automation. The token's account needs automation-administration rights. A successful connection or issue read does not prove that trigger setup will succeed.

Start with a controlled issue review

Choose a dedicated test issue in a project approved for evaluation. Set action permissions before chatting: keep public replies off or approval-gated, and disable other writes that the test does not need. Enable only the necessary lookup actions.

You can then start a named conversation:

npx @eesel/cli new --name "jira-support-review" --agent AGENT_ID
npx @eesel/cli chat 'Review TEST-123 using the connected Jira lookup and available support policies. Summarize the request, identify missing information, and draft a response here. Do not post a comment, assign the issue, or change any records.' --agent AGENT_ID

Replace TEST-123 with your authorized test issue. The prompt asks for a draft in the conversation, not an internal Jira note. That distinction keeps the first check focused on reading and reasoning.

A prompt is not an enforced security boundary. Real chat can invoke enabled tools, so the action settings are essential even when the request says “do not change anything.”

Check whether the draft matches the issue, uses the appropriate policy, and identifies missing facts. If a lookup fails, investigate access and the issue key before treating the generated response as reliable.

Review the result before adding automation

npx @eesel/cli activity --agent AGENT_ID
npx @eesel/cli approvals --agent AGENT_ID

Inspect the work and any held actions. A held proposal has not executed. If you later test an internal note, verify it on the issue itself; a statement in chat is not proof that Jira received the update.

Only then define a narrow automation, such as a specified class of new issues in a test project. Review its enabled state and permissions as part of creation. There is no JSM trigger for an @eesel mention, status change, or arbitrary field change.

The CLI's --dry-run previews the server call a write would make without sending it. It does not run a historical-ticket simulation or prove the quality of a customer response. The CLI documentation covers the exact command options.

How to choose a starting point

Do not enable two automated responders on the same events without deciding who owns each case. An integration existing alongside another product does not establish coordination between them.

For either option, include an inaccessible issue, an ambiguous request, and a policy exception in the evaluation. Your handoff process should name the human responsible when the agent cannot finish.

Try eesel CLI with one Jira support workflow

If terminal access is useful to your team, eesel CLI gives you a concrete way to inspect the support teammate's setup, run an authorized issue review, and check its activity. Your support colleagues can stay in the dashboard and Jira.

Try eesel with a defined Jira support case. Compare the draft, source use, and human review required before expanding to customer-facing automation.