Labsco
MCP SERVER

Agile Planner MCP Server

by cyberlife-coder ↗

Turn a project description into a complete agile backlog — epics, features and user stories as Markdown files — or generate one feature at a time.

Project & Task Management
Summary
The output is a directory of files, not a chat response.

Writing the backlog to disk as annotated Markdown is what separates this from asking a model to draft user stories: the result lands in the repository where the implementing assistant will read it, with checkboxes it can tick. The schema validation step matters too — a backlog that fails structure checks is regenerated rather than written half-formed. `MCP_EXECUTION` is the setting people miss.

What it is

An MCP server that generates agile backlogs through an LLM and writes them to disk as a structured directory of Markdown files. Output goes into a `.agile-planner-backlog` folder with epics containing features containing user stories, plus an `orphan-stories` folder for anything not attached to a feature and a `backlog.json` alongside. Every generated file carries annotations aimed at the AI assistant that will implement the work.

What you get
  • `generateBacklog` — a complete backlog from a `projectName` and a `projectDescription`, with an optional `outputPath`
  • `generateFeature` — one feature from a `featureDescription`, with an optional `storyCount` of at least 3, a `businessValue`, an `iterationName`, an `epicName` to attach it to or let the system pick, and an `outputPath`
  • A validation pipeline rather than raw generation: the structure is checked against a JSON schema, features are enriched with acceptance criteria and tasks, and stories are organised into epics and features before any file is written
  • Task checkboxes for progress tracking and dependency notes in the generated files
  • The same two operations available as a CLI, so a backlog can be generated outside an editor
  • Uniform error handling — every failure from the backlog functions returns the same JSON shape rather than an ad-hoc message
Requirements

An LLM key: `OPENAI_API_KEY` or `GROQ_API_KEY` as the alternative. `MCP_EXECUTION` must be set to "true" for MCP mode — the server will not behave correctly without it. Node.js; the package is `agile-planner-mcp-server` (1.7.3 in package.json) and the documented client entry runs `npx agile-planner-mcp-server`, or `node` against `server/index.js` for a local checkout. `AGILE_PLANNER_OUTPUT_ROOT` sets the base output directory, and `TEST_MODE` produces mock output without calling a model.

Setup effort

One command plus a key — npx agile-planner-mcp-server, then supply credentials