Repository avatar
Other Tools
v0.2.0
active

productops-prioritization

io.github.EugeneNeverov/productops-prioritization

Offline product prioritization: RICE, WSJF, method comparison and capacity what-if scenarios.

Documentation

productops-mcp

npm: prioritization npm: jira CI License: Apache-2.0

Open-source Product Operations toolkit built on the Model Context Protocol: local prioritization (RICE, WSJF, capacity scenarios) and Jira sprint intelligence, plus reusable agent skills. Calculations are done by code; the model only explains the results.

A sprint health audit on the bundled demo data

The tool result is the real output of assess_sprint_health on the synthetic demo sprint; the answer is an abridged version of the skill's example report. Regenerate with node scripts/demo/make-demo-gif.mjs.

Install

Both servers run locally over stdio and need Node.js 20 or newer. Neither needs an account to try: each ships demo data.

Claude Code

claude mcp add productops-prioritization -- npx -y @productops-mcp/prioritization
claude mcp add productops-jira -- npx -y @productops-mcp/jira

Claude Code plugin (both servers and both skills in one install):

/plugin install productops --marketplace eugeneneverov/productops-mcp

On Claude Code older than 2.1.275: /plugin marketplace add EugeneNeverov/productops-mcp, then /plugin install productops@productops. Details in plugin/README.md.

Claude Desktop: download productops-prioritization-<version>.mcpb or productops-jira-<version>.mcpb from the latest release and open it, or add the servers to the config as below.

Any MCP client (Claude Desktop, LM Studio, Cursor and others):

{
  "mcpServers": {
    "productops-prioritization": {
      "command": "npx",
      "args": ["-y", "@productops-mcp/prioritization"],
      "env": { "PRODUCTOPS_DATA_DIR": "/path/to/your/backlogs" }
    },
    "productops-jira": {
      "command": "npx",
      "args": ["-y", "@productops-mcp/jira"]
    }
  }
}

Then ask: "Rank the demo backlog with RICE and WSJF and tell me where they disagree", or "How is the current sprint on the demo PAY board going?". To read your own Jira Cloud, run npx -y @productops-mcp/jira login once and set PRODUCTOPS_JIRA_MODE=live (docs/jira-live.md).

Packages

PackageWhat it is
@productops-mcp/coreShared schemas (Meta, ToolError), tool runtime, policy, config, stderr logger
@productops-mcp/prioritizationLocal MCP server (stdio), fully offline: calculate_rice, calculate_wsjf, compare_initiatives, simulate_capacity, explain_ranking, optional triumvirate_brief / triumvirate_verdict (three independent judges)
@productops-mcp/jiraRead-only MCP server (stdio) for sprint health: list_projects, get_sprint_issues, analyze_cycle_time, find_blockers, assess_sprint_health. Ships a synthetic demo dataset; reads your Jira Cloud through the official Atlassian MCP server after productops-jira login (docs/jira-live.md)

Try it from a checkout

After pnpm install && pnpm build, ask the server to rank the bundled demo backlog:

pnpm client packages/mcp-prioritization/dist/bin.js calculate_rice '{"initiatives":{"kind":"demo"}}'

And audit the demo team's sprint (synthetic Jira data, no account needed):

pnpm client packages/mcp-jira/dist/bin.js assess_sprint_health '{"boardId":1}'

Your own backlog goes in a CSV or JSON file inside an allowed folder: PRODUCTOPS_DATA_DIR, ./data, or the folders in productops.config.json. The columns match the fields in docs/algorithms.md; the sprint metrics are defined in docs/metrics.md. Sample outputs are in examples/reports/demo.

Skills

skills/roadmap-prioritization turns the tools into a recommendation a product manager can act on: what to take, what to defer, the trade-offs, where the methods disagree, and how to reproduce every number. It is also served as the MCP prompt roadmap-prioritization, so it works in clients without Agent Skills support, such as LM Studio. skills/sprint-health-audit does the same for a sprint (MCP prompt sprint-health-audit). See skills/README.md and an example report.

Skill quality is measured by evals: 10 cases, deterministic checks (every number and id must come from a tool), and a runner for any OpenAI-compatible endpoint, LM Studio included.

Development

Requires Node.js ≥ 22.12 and pnpm 10 (corepack enable). Published packages run on Node ≥ 20.

pnpm install
pnpm test          # build, then unit + e2e tests
pnpm lint && pnpm format:check
pnpm client        # start the built server and call `ping` through a real MCP client

To connect a client to your checkout instead of npm, see examples/clients. Design decisions are recorded in docs/decisions. Releases are described in docs/releasing.md and listed in CHANGELOG.md.

License

Apache-2.0, see LICENSE. This license covers the code in this repository only. Use of third-party APIs (Atlassian Jira, Google Analytics and others) is governed by those providers' own terms.

NPM
@productops-mcp/prioritization
Install Command
npm install @productops-mcp/prioritization
Runtime: npx