Release notes in the customer's words

Product and designPM writing a monthly customer changelog with Claude Code

Every month I asked the agent to turn merged work into release notes. It used internal words like 'tenant' and 'ingest worker' and asked engineers what each change meant, again.

How do I get Claude Code to write release notes in customer words instead of internal jargon?

Save your customer-facing names once with ArcRoom's remember, like 'workspace' not 'tenant'. Engineers' agents save explanations of changes with complete_task, and your agent finds them through find_task, then reads git log for the rest. Save wording choices with arcroom decide. ArcRoom doesn't store diffs, so changes nobody explained still need a question to the engineer.

Also asked as: My AI-written changelog keeps using internal terms like tenant. How do I fix it? · How do I write a monthly changelog without asking engineers what each change means? · Can my agent reuse change explanations that engineers' agents already wrote?

The work

Customer-facing names are saved once with remember: 'workspace' not 'tenant'.

ArcRoom keeps
  • Reused answer
  • Saved answer
  • Decision
  • Saved decision
The next agent gets

Remember: customer words: workspace (not tenant) · import (not ingest) · no internal ticket ids

Step by step

  1. Customer-facing names are saved once with remember: 'workspace' not 'tenant'.
  2. Engineers' agents already save explanations of changes with complete_task.
  3. The PM's agent finds those explanations through find_task and reads git log for the rest.
  4. Wording choices are saved with arcroom decide.

What the agent sees

Remember: customer words: workspace (not tenant) · import (not ingest) · no internal ticket ids
find_task → 4 saved explanations this month (export limit, SSO fix, …)

Compared with today's tools

git log + a glossary doc

Paste the month's git log and a glossary into the agent.

Setup
1 hour for the glossary
Ongoing
Paste both each month, chase engineers, ~1 h

ArcRoom

Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.

Setup
2 min (npx arcroom init)
Ongoing
None

Mem0 can hold the glossary if saved, but the engineers' explanations only exist if their agents chose to save to Mem0; in ArcRoom they're already there from complete_task.

Where it stops

ArcRoom doesn't store diffs or code. Changes nobody explained still need git log and a question to the engineer.

Try it in your project

Works with the agents your team already uses, and any agent over MCP or the HTTP API.

More in product and design