Thanks for contributing to agent-memory.
npm ci
npm run build
npm run testUseful local smoke tests:
npm run smoke:consumer
npm run smoke:real-provider
node dist/cli.js init --yes
node dist/cli.js status
node dist/cli.js query "what should I do next?"
node dist/cli.js integrate --status
node dist/cli.js integrate --dry-runPackage-level smoke test:
npm run pack:dry-runFor command flows that rely on Codex or Claude, agent-memory now probes provider availability before the command does deeper work. If a provider is missing or not authenticated, fix that first rather than debugging the later command output.
The public npm package is @agent-connect/memory, while the CLI command remains agent-memory.
Typical consumer workflow:
npm install -D @agent-connect/memory
npx agent-memory init
npx agent-memory status
npx agent-memory query "what should I do next?"
npx agent-memory integrate --status
npx agent-memory integrate --dry-run
npx agent-memory mcp- Keep the tool small and explicit.
- Prefer conservative behavior over surprising automation.
- Treat human-edited project memory as higher priority than generated convenience.
- Keep README and docs aligned whenever command behavior changes.
Good changes usually include at least one of these:
- clearer command behavior
- safer ownership and merge rules
- better generated memory quality
- better documentation for adoption and maintenance
If a change affects init, update, or validate, include the relevant CLI smoke test in your local verification notes.