memclaw-maintancev0.9.31

A developer or OpenClaw operator installs the @memclaw/memclaw plugin, enables it in openclaw.json, disables the legacy memorySearch, and wires up LLM and embedding API keys. They then restart the Gateway and verify Qdrant on 6333/6334 and cortex-mem-service on 8085.

An existing OpenClaw user with native memories in memory/*.md and MEMORY.md wants to preserve their history. They run cortex_migrate once to convert files into MemClaw L2 format and generate L0/L1 layers and vector indices.

A user replaces the legacy Memory section in AGENTS.md with MemClaw-specific session startup, recording, searching, and profile-building instructions. This ensures the agent stops using MEMORY.md and daily logs and adopts cortex tools consistently.

An operator runs cortex_maintenance with commands like prune, reindex, and ensure-a to keep the Cortex memory store healthy. They may first run with dryRun:true to preview changes before committing.

After restarting OpenClaw, an admin checks whether Qdrant and cortex-mem-service are reachable via their health endpoints. If a port fails, they inspect plugin configuration and API key completeness.

The @memclaw/memclaw plugin is free and open source, but users bring their own LLM and embedding API keys. The provider monetizes indirectly through API usage consumed by memory extraction and semantic search.

A hosted version of MemClaw bundles Qdrant, cortex-mem-service, and API key management so users avoid local setup. Subscription tiers differentiate storage, retention, and support.

Organizations deploying OpenClaw at scale pay for assisted installation, AGENTS.md policy templates, migration from legacy memory, and ongoing maintenance guidance. Support includes SLA-backed troubleshooting.