Terms
One word for one thing. Your agents keep to it.
Terms are your project's glossary: one word for each thing in your product, with its definition, checked by your coding agents before they write. They give the context layer one vocabulary for your team, your docs and every agent.
Why agents invent words
A coding agent starts each session without your project's words, so it picks its own: “workspace” in one session, “team” in the next, “account” in a third, all for the same thing.
Each word lands in the code, the database or the interface, and once it has spread, fixing it is a migration, not an edit. Terms give every agent your word before it writes its own.
Terms with a definition
A term is one word of your project's vocabulary with a one-line definition, its other names and the words it must not be confused with: the ubiquitous language of domain-driven design, kept where your agents look it up.
Every other name leads to the term. A new word works at once, as a draft until a person makes it the project's word, and a retired one names the word to use instead.
Checked before agents write
Before an agent writes, it looks up the words it means to use and runs the term check on its draft, so a wrong word is caught in the draft, not in the code.
The term check flags every former or retired word with the one to write instead, in a draft, a page or the whole project. An agent that needs a word your project lacks proposes it as a draft term.
Rename once, everywhere
Renaming a term changes its word everywhere the term is used and keeps the old word as a former name, so the old word still finds it.
Every tag and link shows the new word at once, and nothing breaks. The term check flags text that still uses the old word.
Definitions where words appear
The first time a docs page uses one of your terms, a hover card shows its definition, and records and the terms themselves do the same.
Nobody writes a link for it, and a term you add later explains text written before it.
The same words for every agent
Every agent that connects to your project knows your terms are there and looks words up as it works, so Claude Code, Codex, Cursor, Gemini CLI and Grok work from the same vocabulary.
Each looks up only the words its task uses, so none spends context on the rest.
Renaming “workspace” to “organization”
Your product first called its accounts workspaces. Now the word is organization, and your agents keep writing the old one.
Rename the term
You rename “workspace” to “organization”. The old word stays as a former name, and you add “org” as an alias and “project” as the term not to confuse it with.
Tags follow at once
Every file, page and item tagged with the term shows “organization” from that moment.
The agent looks it up
An agent writing the invitation emails looks up “workspace”, gets “organization” back as the term it names, and writes that.
The term check finds the rest
A term check over the project flags the old word in a runbook and a decision record, each with “organization” beside it, and the agent fixes both.
The code, the docs and the interface say “organization”, whichever agent wrote them.
Questions about terms
What is a term, and how does it relate to a ubiquitous language?
How do my coding agents use terms?
What about synonyms?
What happens when I rename a term?
Does it work with any coding agent?
Give your project one vocabulary
Start with the words your product already uses. Your agents check against them from their next session.