Quick start

Run an agent turn

Every command is scoped to an organization and project and writes to a local memory database. Set those four global flags before the agent subcommand.

export ANTHROPIC_API_KEY=sk-ant-...

codel00p \
  --memory-db ./codel00p.db \
  --organization-id acme \
  --project-id web \
  --project-name "Web App" \
  agent run "Summarize how this project is built and tested" \
  --provider anthropic \
  --model claude-opus-4-8 \
  --tool-set all

--tool-set selects which tools the run may use: read, edit, command, git, or all. Add --session-id to persist the session so you can resume it later with agent resume.

Review the memory it proposed

After useful work the run proposes memory candidates. List them, then approve the ones worth keeping. The four scope flags are the same as above.

codel00p ...scope memory list
codel00p ...scope memory approve <memory-id>

Only approved memory reaches the model on future runs. Next, the Tutorialwalks through the interactive chat, the way you'll actually work day to day.