Publish Your Changelog From Claude Code

Brian Sierakowski - Founder

Claude Code is becoming a natural place to ask, “What did we just ship?”

That question should not end with a paragraph trapped in your terminal. It should end with a reviewed changelog entry, a durable URL, and the right version routed to the right audience.

That is the point of the Changebot MCP workflow.

The basic flow

Start with the question you already want answered:

Ask Changebot: write a customer-facing changelog update for what shipped last week.

Changebot can return a draft, the source context it used, and a review URL:

Draft ready: "Billing exports are easier to reconcile"
Review: https://app.changebot.ai/review/billing-exports
Suggested destinations: hosted changelog, customer widget, internal Slack

Then you keep working in the same surface:

Make it less technical, add a support-team version, and publish the customer version.

The important part is not the exact command syntax. The important part is the shape of the workflow: ask, review, edit, approve, publish.

Why MCP changes the changelog workflow

Most changelog tools assume you will leave your work, log in, and write inside their interface.

MCP flips that. The agent surface becomes the starting point. Claude Code, ChatGPT, Slack, or another MCP-compatible client can collect intent and source material. Changebot turns that into a publishable update and sends it to the changelog destination.

That matters because the source material is already scattered:

  • The actual code lives in GitHub or Bitbucket.
  • The planning context lives in Linear, Jira, or Slack.
  • The draft might start in Claude or ChatGPT.
  • The final customer-facing record needs to live on a URL.

Changebot connects the last mile.

What happens after the draft

A generated release note is not done just because the words exist.

Someone still needs to decide whether the update is customer-facing, whether it should be internal-only, how much detail support needs, and where the final version should live.

The MCP workflow keeps those decisions explicit:

  1. The agent drafts the update.
  2. Changebot returns a review link.
  3. A human edits or approves.
  4. Changebot publishes to the hosted changelog, widget, Slack, API, or custom domain.

That is different from raw summarization. Raw summarization gives you text. Changebot gives you a publishing workflow.

Start with one update

You do not need to connect every repo on day one.

Start with one update. Paste a rough launch note, ask for a customer version, review the draft, and publish it to your hosted changelog. When that workflow works, add GitHub or Bitbucket so Changebot can draft from what actually shipped.

Start free, or read more about the Changebot MCP.