CLI Reference

Decisions & memory

intent decide, note, and example — building the Memory and Examples layers of the context harness.

These commands populate the Memory and Examples layers of the context harness — the parts of CLAUDE.md that keep an agent from re-litigating a decision you already made, or re-deriving a pattern you already established.

intent decide [title]

intent decide
intent decide "Use Postgres row-level security for multi-tenancy" \
  --description "Simpler than an app-layer tenant filter and harder to accidentally bypass" \
  --category architecture \
  --story a1b2c3

Any field you don't pass on the command line is prompted for interactively. Logging a decision here is what makes it show up in CLAUDE.md after intent sync --context — so the next agent session (or the next teammate) doesn't have to ask "wait, why did we do it this way?"

intent note <storyId> [note]

intent note a1b2c3
intent note a1b2c3 "Auth redirect bug was a stale cookie domain, not the token itself"

Adds an implementation note to a story — prompts for the text if you don't pass it. Updates both the story on intentdocs.com and the local .intent/stories/<storyId>.md context file, prepending a timestamp under that file's "Implementation notes" section (this is one of the sections intent sync preserves rather than overwrites).

intent example

intent example

Interactively prompts for a few-shot reference pattern: title, description, language (typescript, javascript, python, bash, sql, or other), and before/after code (multi-line — enter a blank line twice to finish each block). Saves it to the project's Examples. Run intent sync --context afterward to regenerate CLAUDE.md with the new pattern included — the command reminds you of this.