Browse documentation

Connect AI agents

CLI adapter

Use Memory through command-line workflows.

On this page

Use this adapter when an agent can run local shell commands but does not have Zharwing Memory MCP tools.

Startup#

bash
zharwing-memory detect <working-directory>
zharwing-memory status --project <project-id>
zharwing-memory resume --project <project-id>
zharwing-memory search --project <project-id> "<task, feature, error, or file>"

Start a session when the work is meaningful:

bash
zharwing-memory start "<task title>" --project <project-id> --agent <agent-name>

Attach a known workstream when appropriate:

bash
zharwing-memory start "<task title>" --project <project-id> --agent <agent-name> --workstream <workstream-id>

Context#

Preview context before using it in a prompt:

bash
zharwing-memory context --project <project-id> --preview --task "<task>"

Persist the bundle when it is actually used:

bash
zharwing-memory context --project <project-id> --session <session-id> --task "<task>"

Progress#

bash
zharwing-memory checkpoint --project <project-id> --session <session-id> "summary" \
  --next "next step" \
  --file "path/to/file"

Close when the work round is done:

bash
zharwing-memory close --project <project-id> --session <session-id> "summary" \
  --next "next step"

Agent Instructions#

Generate project-specific instructions from the neutral protocol:

bash
zharwing-memory agent-instructions --project <project-id> --agent generic
zharwing-memory agent-instructions --project <project-id> --agent codex --output AGENTS.md
zharwing-memory agent-instructions --project <project-id> --agent claude --output CLAUDE.md
zharwing-memory agent-instructions --project <project-id> --agent qwen --output QWEN.md

Generated files are adapters. They are not the source of truth.