Skip to main content
Glama

Context Pack

Know a Jira work item is ready, then hand its full context to a person or an AI coding agent.

AI coding agents are only as good as the context they get. A ticket that says "Make checkout faster" gives them nothing. The real context is spread across the work item, its links, a Confluence spec, decisions buried in comments, the goal it serves and the people who know the area. Context Pack gathers all of it in one place, checks it against a Definition of Ready, and serves it to every surface a team uses:

Surface

What it does

Jira issue panel

Readiness score with a one-line fix per gap, plus a Context tab and a paste-ready handoff for agents

Readiness board (project page)

The top 25 open work items of the space's backlog, most ready first, with what each one is missing

Definition of Ready (project settings)

Project admins choose which checks apply, the minimum description length and how strict "ready" is

Rovo agent "Context Pack"

"Is this ready to start?", "Draft acceptance criteria", "Which items are ready?"

Rovo skill ready-for-dev

Reusable instructions that take a work item from "is it ready?" to a plan

MCP server (rovo:mcp)

The same four tools for Claude Code, Cursor and any MCP client, with the user's own Jira permissions

What goes into a pack

The pack covers the six kinds of context a team runs on:

  • Work: summary, type, status, priority, owner, parent, linked work items and open blockers (Jira REST).

  • Knowledge: linked Confluence spec pages, including their text, owner and last update (Teamwork Graph).

  • People: the assignee, the reporter, and the spec owners and contributors to ask (Jira and Teamwork Graph).

  • Communication: decisions and out-of-scope notes found in the comments.

  • Code: pull requests and deployments linked to the work item (Teamwork Graph).

  • Assets: designs linked to the work item (Teamwork Graph).

  • Goals: the goal the work contributes to, which is the "why" an agent usually never sees (Teamwork Graph).

Related MCP server: jira-mcp-server

Definition of Ready

Six explainable checks, with no black-box score:

  1. Has a clear description (at least 20 words)

  2. Has acceptance criteria (an "Acceptance criteria" section, or Given / When / Then lines)

  3. Is estimated (story points or an original estimate)

  4. Has an assignee

  5. No open blockers (an "is blocked by" link to a work item that is not done)

  6. Links a spec or design page (a Confluence remote link, or a wiki link in the description)

By default an item is ready when all checks pass, or when acceptance criteria and blockers pass and at most one other check fails. The score is the share of passed checks.

Project admins can change this per space under Project settings → Context Pack: switch checks off (for example estimates, for teams that do not estimate), set the minimum description length, or require every check to pass. The panel, the board, the Rovo agent and the MCP tools all use the same settings.

Tools (Rovo actions, also exposed over MCP)

Tool

Verb

Inputs

Returns

get-context-pack

GET

issueKey (optional on a work item)

readiness summary + the full pack as markdown

check-readiness

GET

issueKey (optional on a work item)

score, ready flag, every check with its fix

list-ready-items

GET

projectKey, limit (max 25)

open items, most ready first, with their gaps

post-progress

CREATE

issueKey, message

adds a "Progress update" comment

Some MCP clients show the tools with the app environment appended, for example check-readiness_production.

How it is built

  • Forge app, Runs on Atlassian (no egress, no external services, no paid AI). Node 24 runtime, UI Kit only.

  • Every Jira call runs as the current user (asUser), so nobody sees anything they could not open in Jira.

  • Teamwork Graph: curated queries (jiraWorkItemGoals, jiraWorkItemLinkedExternalPullRequests, jiraWorkItemLinkedExternalDeployments, jiraWorkItemLinkedExternalDesigns, confluencePageOwners, confluencePageContributors) plus one persisted Cypher query that returns a spec page by its ARI:

    MATCH (page:ConfluencePage {ari: $pageId}) RETURN page

    Graph calls are fail-soft: if the graph is unavailable, the pack is still built from Jira data and says what is missing.

  • Per-project settings live in Forge Key-Value Storage (storage:app). Saving them requires the Jira "Administer projects" permission, checked as the current user.

  • Pure rules (parsing, readiness, markdown) live in src/lib/rules.js and are unit-tested with Node's built-in test runner (npm test, 16 tests).

src/
  lib/rules.js            Definition of Ready, ADF parsing, markdown handoff (pure, tested)
  lib/contextPack.js      Jira REST reads and the pack builder
  lib/teamworkGraph.js    Teamwork Graph queries
  lib/settings.js         per-project Definition of Ready (Forge storage)
  actions.js              Rovo action / MCP tool handlers
  resolvers/              UI resolvers (panel and board)
  frontend/               UI Kit panel and readiness board
skills/ready-for-dev/     rovo:skill instructions
agent-skill/              the same workflow as an open skill for third-party agents
test/rules.test.mjs       unit tests

Run it yourself

npm install
npm test                                  # unit tests
forge deploy -e development
forge install --site <your-site>.atlassian.net --product Jira -e development

To let a coding agent use the tools over MCP, a site admin opens Atlassian Administration → your site → Connected apps → Context Pack → Details → Tools and selects Enable tools. Then, for example in Claude Code:

claude mcp add --transport http context-pack https://mcp.atlassian.com/forge/e9f46537-efa3-4c85-99f5-9bec166d7368

Run /mcp to sign in with your Atlassian account and pick your site.

Privacy and terms

Context Pack runs on Atlassian Forge, calls no external services and stores no personal data. See PRIVACY.md and TERMS.md.

Questions and bug reports: GitHub issues.

License

MIT

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      B
      quality
      D
      maintenance
      Provides tools for AI assistants to interact with JIRA APIs, enabling them to read, create, update, and manage JIRA issues through standardized MCP tools.
      6
      10 npm
      3
      MIT
    • A
      license
      A
      quality
      D
      maintenance
      Enables Claude AI and other MCP clients to interact with Jira Server/Data Center through tools like listing issues, logging work, and updating issues, requiring user confirmation for write operations.
      8
      1,667 npm
      1
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      A Model Context Protocol server that connects your AI IDE to Jira. Query your tickets, active sprint, and issue details directly from GitHub Copilot, Cursor, Claude Desktop, or any MCP-compatible client.
      21 npm
      ISC
    • A
      license
      Not graded
      quality
      D
      maintenance
      Wraps Jira Data Center operations as MCP tools, enabling AI assistants to interact with Jira for project management and issue tracking.
      1,667 npm
      Apache 2.0