A glossary for every language
Research and writingTeam translating a product into Hindi and Spanish
One agent translated 'workspace' as one word, the next session used another. Users saw three different words on the same screen.
How do I keep AI translations into Hindi and Spanish using the same terms across sessions?
Save your glossary choices as decisions in ArcRoom, and every translation prompt gets the glossary in recall. Terms then stay the same across files and sessions. A large glossary may not all fit in one recall, but the agent can still ask find_task.
Also asked as: Our app shows three different words for one term after AI translation. How do we fix that? · How do I give an AI translator a glossary it keeps using? · Can an AI agent remember how we translated a term last time?
Glossary choices are saved as decisions.
- Decision
- Recall
Decision: Hindi — 'workspace' → 'कार्यक्षेत्र', 'memory' → keep English.
Step by step
- Glossary choices are saved as decisions.
- Every translation prompt gets the glossary in recall.
- Terms stay the same across files and sessions.
What the agent sees
Decision: Hindi — 'workspace' → 'कार्यक्षेत्र', 'memory' → keep English.Compared with today's tools
Glossary file in the repo, linked from rule files
Keep a term list per language in the repo and tell every agent to use it via CLAUDE.md and AGENTS.md.
- Setup
- 30 min
- Ongoing
- add terms as they come up
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
A checked-in glossary is the more reliable source of truth; ArcRoom recalls only matching terms, which helps when the list is too long to load every time.
Where it stops
Large glossaries may not all fit in one recall. The agent can still ask find_task.
Try it in your project
Works with the agents your team already uses, and any agent over MCP or the HTTP API.