AI Quick Start

Prefer to read the docs yourself?

If you want to go through Module Federation concepts, configuration, and integration in order, head to the Getting Started guide.

A Better Way To Start

If you are already using an AI coding tool, the better starting point is not to read all of the docs yourself first.

The faster approach is: give the latest Module Federation docs to your agent first, then let it answer questions, debug problems, update config, and write code.

Many models still have incomplete or outdated knowledge of Module Federation 2.0. So the point of this page is not to make you study first. It is to help your agent get the right source material first.

If you use Claude Code, Cursor, Windsurf, or another editor with Skills support, install:

npx skills add module-federation/agent-skills --skill mf -y

Then ask the agent to read the docs before answering:

/mf docs What's the difference between singleton and requiredVersion in shared?

The idea is simple:

  • you do not need to hunt through the docs first
  • your agent reads the latest docs before it answers
  • the result is much more likely to be accurate

If You Want Your Agent To Do More

The mf skill does more than answer questions. It can also help your agent:

  • add Module Federation to an existing project
  • inspect config problems
  • analyze type issues
  • check shared dependency conflicts
  • inspect remote module information
  • use the Divebell MF Extension to diagnose runtime failures, loading chains, and performance bottlenecks
  • read application observability reports from local, SSR, or production environments

👉 View Agent Skills documentation

Observe and Debug MF Loading

Remote resolution, Shared selection, Bridge lifecycle, and module performance problems often exist only in the live page runtime. For a one-off diagnosis, the mf skill can use the Divebell MF Extension to obtain bounded, structured evidence without adding the Observability Plugin to the application. Page interaction is used only to trigger the loading path.

Ask the agent directly:

/mf observability
Open https://example.com with Divebell, reproduce the failing checkout flow,
and tell me which Remote or Shared phase failed.

When needed, the skill installs Divebell globally on the agent machine, reads the CLI Skill returned by divebell skill, installs @divebell/extension-mf, and reads the command Skill returned by divebell mf --skill. Neither package is added to your application.

For continuous collection, Node/SSR files, production uploads, or reports that must be retained, integrate the Observability Plugin. After a loading failure, give the generated traceId, read: command, or uploaded report to the agent:

/mf observability
I saw this Module Federation console error:

[Module Federation] Observability report generated
traceId: mf-...
read: window.__FEDERATION__.__OBSERVABILITY__["runtime_host"].getReport("mf-...")

Please read the report and fix the issue.

For Node or SSR, ask the agent to read .mf/observability/latest.json. For production, upload reports with the plugin's onReport callback and give the stored report to your agent. A trace ID from a closed browser session is not a substitute for the stored report.

You Can Still Use This Without Installing Anything

If you just want a lightweight version first, send this directly to your agent:

Before answering, please read https://module-federation.io/llms.txt to learn about the latest Module Federation 2.0 documentation, then answer my question.

This is the easiest way to stop the model from guessing when you have not installed the skill yet.

Docs You Can Hand To Your Agent

llms.txt

module-federation.io supports the llms.txt standard, which is a good entry point for AI systems:

https://module-federation.io/llms.txt

Markdown Docs

Every page on the docs site is also available in Markdown. Replace .html at the end of a docs URL with .md:

# HTML version
https://module-federation.io/guide/basic/runtime.html

# Markdown version
https://module-federation.io/guide/basic/runtime.md

Next

If you want to start using this workflow right away:

  1. Install mf
  2. Ask the question you care about most
  3. Use Divebell for one-off MF runtime, loading-chain, or performance diagnosis
  4. Enable the Observability Plugin when loading evidence must be retained or uploaded
  5. Open Skills documentation to see which sub-commands it supports