Agent Ready for Jira – User Guide

How to use Agent Ready: checks on every issue, the Backlog X-ray, project rules, the handoff gate, Rovo, MCP tools for coding agents and the AI agents report.

Get started in 2 minutes

  1. Install Agent Ready from the Atlassian Marketplace (a Jira admin; free for up to 10 users). Every project is checked at once with the Balanced rules — nothing to set up.
  2. Open the Backlog X-ray: Jira’s Get started for Agent Ready, or Jira settings → Apps → Agent Ready → Backlog X-ray. It shows how many open issues are ready and which are one check away.
  3. Open any issue: Agent readiness in the right column says what is missing.
  4. Press Fix on an issue one check away, add the missing section and Insert into description — the issue is checked again.

Agent readiness on every issue

Agent readiness sits in the issue’s right column.

  • Collapsed, its lozenge shows the last check: Ready, a score such as 5/8, Not checked (nobody opened it since the rules changed) or Not set up.
  • Checks lists what is missing and what to add; passed checks fold into one line. Check again after you edit the issue.
  • Copy missing sections copies the headings your rules look for. Draft of the missing sections starts with your project’s headings: write under them (or paste Rovo’s draft) and press Insert into description — only the sections the issue misses are added, at the end, as you.
  • Write a criterion step by step turns three short answers — Given, When, Then — into one acceptance criterion. No AI.
  • Split into sub-tasks drafts one sub-task per criterion for an issue too big for one agent run.
  • For the agent: the roles and skills chosen for the issue and the brief the agent gets; Copy brief for the agent.
  • Waive a check that is not needed on this issue from its ⋯ menu, with a reason; Undo brings it back. Who may waive is a project setting; passwords and hidden instructions only ever by project admins.

The checks read the issue as you see it: nobody sees more through Agent Ready than in Jira.

The checks

CheckPasses when
Descriptionthe description has at least the minimum number of words
Acceptance criteriaa story or task has an acceptance criteria heading with something under it
Steps / expected / actuala bug has steps to reproduce, the expected and the actual result
Small enoughthe estimate is within the project’s story points limit
No open blockersno “is blocked by” link to an issue that is not done
Linka link to a design, spec or file (off by default)
Agent role chosena role is chosen for the issue (off by default)
No passwords or keysnothing that looks like a password, API key or token in the summary, description, links or latest 50 comments
No hidden instructionsalways on: no text that tries to give an AI agent orders
Repository knownthe project maps the issue to a repository (on in Strict for agents)
Clear wordingno vague words, and every criterion is testable (on in Strict for agents)
One agent runnot more criteria or words than the limits (on in Strict for agents)

Your own required sections count as checks too.

Passwords and keys: about 50 kinds of keys and tokens, password= lines and JWTs. Only the type and a mask (ghp_****q7R8) are shown, never the value, and nothing of it is stored. Remove the value from the issue and change it.

Hidden instructions: orders to ignore the agent’s instructions, role switches, requests to send secrets or run what a link says, invisible characters, hidden HTML comments and long encoded blocks. A portal customer’s text is checked more strictly. The agent never gets a flagged line from Agent Ready.


Backlog X-ray

Jira settings → Apps → Agent Ready → Backlog X-ray (Jira admins) checks the 100 most recently updated open issues in every project you can see, each with its project’s rules:

  • one sentence — “8 of 33 open issues are ready for AI agents” — what is missing most, how many are one check away, and the trend against a week before;
  • One check away: the issues to fix first, each with Fix in the row;
  • Rules for every project: Light, Balanced or Strict for agents in one click;
  • Copy for the standup: the X-ray as plain text to paste anywhere.

An empty site offers sample data there.


Rules

Rules for every project (Jira admins)

Jira settings → Apps → Agent Ready → Rules for every project: the Brief for every project — your company’s rules for every agent — and the rules used by every project without rules of its own. Check projects that have no rules of their own with these is on from the install.

A project’s rules (project admins)

Project settings → Agent Ready → Readiness rules starts with Setup, a list of what is done and left, then seven tabs and one Save:

  1. Issue types — how each type is checked: as a story, as a bug, rules only, or not checked; its own sections; the rules that differ on it.
  2. Rules for every type — start from a preset, then switch rules on or off.
  3. Agent brief — what every agent must know that is not in the code: Definition of Done, what not to touch, how to name pull requests. Skills in every brief go with it.
  4. Roles — Security, Backend, QA… each with its own brief, skills and optionally its own agent.
  5. Repository — a default repository and base branch, and rules by component, label or issue type; the first line of every brief names it.
  6. Agents — which agent accounts may take this project’s issues, plan approval for risk areas, the review limit and monthly budget.
  7. Headings — the words your team writes for each section, in any language.

Preview shows the checks a sample issue would get with the rules as they are on the page, before you save.

Agent skills

How-tos for agents — “Playwright tests”, “DB migrations” — in Markdown. Jira admins keep the company’s (Jira settings → Apps → Agent Ready → Skills); project admins keep a project’s own and turn off company skills that do not fit. Import SKILL.md, .zip or .tgz reads the files in your browser; Copy as SKILL.md and Download SKILL.md take a skill to a repository.


The project’s Agent readiness page

In the project’s navigation, Agent readiness shows at a glance how many open issues are ready, what they miss most and how many are one check away, then:

  • Issues: all open issues, This sprint or Next sprint;
  • Agent-suitable: ready issues nobody holds; choose some, choose an agent and Hand to agent;
  • the open issues Not ready (or All) with a readiness bar and what each misses;
  • Make them ready with Rovo: the ten least ready issues for the Rovo agent (where Rovo is on);
  • the AI agents tab: the report for this project.

Handing issues to agents

  • The “Issue is agent-ready” validator: add it to a transition such as Ready for agent (company-managed: the transition’s Validators; team-managed: its rules). An unready issue stays where it is, and Jira says what is missing.
  • Readiness in JQL: agentReady = "true", agentReadyScore < 80, agentRole = "Backend", agentSkill = "playwright-tests" — kept fresh on every edit.
  • AI agent accounts (Jira settings → Apps → Agent Ready → Settings): find each agent by name; app accounts such as Claude, Copilot and Cursor are marked.
  • Give unready issues back (off by default): an unready issue assigned to an AI agent goes back to the previous assignee with a comment naming what is missing.
  • The brief at handoff (a project setting, off by default): the agent gets the company’s and the project’s brief as a comment.
  • Warn when a sprint starts (on by default): Agent Ready comments on each unready open issue of a sprint that starts, at most 20 a sprint.

Ask Rovo

On sites with Rovo, the Agent Ready agent answers “Is this issue ready for an AI agent?” and drafts the missing sections from the issue’s own text and your agent brief. Passwords and keys reach it masked; lines that look like instructions to an AI agent never reach it. When you agree, it adds the drafts to the description (only the missing sections) or creates sub-tasks — Rovo asks you to confirm first, and it acts as you. It also answers “Is the next sprint ready for agents?”, “What can agents take today?” and “How did agents do this month?”.


Coding agents through MCP

Agent Ready offers its own tools on Atlassian’s MCP server, so a coding agent works from Jira as the person it acts for:

ToolWhat it does
check-readinessready or not, what is missing, whether to start
get-ready-specthe issue as checked: repository, summary, description, comments — keys masked, hidden instructions removed
get-agent-briefthe brief: repository, the company’s, project’s, type’s, roles’ and skills’ instructions
get-skillone skill’s instructions
next-ready-issuethe best issue an agent can take now
start-issuethe gate (an unready issue is refused), a claim, the spec and the brief
ask-questiona question for the team, shown on the issue and posted as a comment
get-answersthe answers people gave on the issue
submit-planthe plan before any code; a risky issue waits for a person’s approval
get-plan-decisionapproved, sent back with a note, or waiting
report-progressone step of the run, shown on the issue
finish-issuethe receipt — what changed, the tests, which criterion each test proves — and the move to review
release-issuestop without finishing: the claim released, the reason in a comment

Turn them on (a site or organization admin, once): admin.atlassian.com → Apps → Sites → your site → Connected apps → Agent Ready. The agent signs in as you (OAuth). Then add the rule from Settings → Setup → Copy the rule to your repository’s AGENTS.md, CLAUDE.md or .cursor/rules.

While an agent holds an issue, Agent work on the issue shows its questions (answer them there), its plan (Approve or Send back…) and its steps, live. Every receipt is checked against the issue’s development information in Jira: Verified, Unverified or Contradicted.


AI agents report

AI agents — on the project’s page, and for every project under Jira settings → Apps → Agent Ready — shows, for a period, one agent or one issue type: issues handed to agents, stopped by Agent Ready, done the first time, given back, reopened and the median time to done; the share done the first time by readiness at handoff; what given-back issues cost once a Jira admin sets the price of a run; the receipts and the criteria they proved; and each agent’s tool use. Learned from give-backs offers the rules your given-back issues missed. Copy as CSV; the Audit trail copies every handoff as CSV or JSON.

The report starts the day Agent Ready is installed. Two dashboard gadgets show a project’s readiness and the agents’ results.


Licence

Without an active licence Agent Ready checks nothing and stops no handoff: the pages say Agent Ready needs a licence, the lozenge says No licence, and the workflow validator lets every transition pass so your workflows keep moving.

Privacy and data

Agent Ready runs on Atlassian Forge and makes no calls outside Atlassian. What it keeps, and for how long, is in the privacy policy.