feat(integrations): add Claude and Codex host bridges

[fryeggs]

Summary

This is PR 4/4 in a series that extends memory-lancedb-pro in layered steps.

This PR proposes the host-integration layer that connects the shared runtime to real user-facing hosts.

It adds bridges for:

• Claude Code
• Codex CLI
• Codex Desktop
• Claude Desktop

Why

A practical limitation of a host-native memory plugin is that real work no longer lives in one host.

In day-to-day usage, memory often needs to flow across:

• OpenClaw
• Claude Code
• Codex CLI / Desktop
• Claude Desktop

This PR shows how the same LanceDB backend can be reused across those hosts instead of splitting memory into isolated silos.

Scope note

This is the largest proposal in the series.

It is intentionally submitted separately from the smaller retrieval / capture / runtime PRs so it can be reviewed independently. Conceptually, it builds on the standalone runtime introduced in PR 3/4.

Validation

npm ci
node --check host-runtime.mjs
node --check mcp-server.mjs
node --check host-integrations/claude-code/stop-capture.mjs
node --check host-integrations/claude-code/user-prompt-submit.mjs
node --check host-integrations/claude-desktop/controller.mjs
node --check host-integrations/claude-desktop/main-hook.cjs
node --check host-integrations/claude-desktop/preload.cjs
node --check host-integrations/codex/notify.mjs
node --check host-integrations/codex/stop-capture.mjs
node --check host-integrations/codex/user-prompt-submit.mjs

All passed locally.

[AliceLJY]

Thanks for this great contribution! 🙏

This PR currently has merge conflicts with main. Could you please rebase against the latest main?

We noticed you have a series of related PRs (#292, #293, #294, #295) — if there are ordering dependencies, let us know the preferred merge order so we can coordinate.

[rwmjhb]

Closing — this PR adds multi-host integration bridges (Claude Desktop Electron hooks, Codex CLI bridges, MCP server) totaling 6000+ lines. This is a significant scope expansion beyond the core OpenClaw memory plugin.

We'd recommend maintaining this as a separate integration package (e.g. memory-lancedb-pro-host-bridges) so the core plugin stays focused on memory lifecycle. A standalone package would also let you iterate on host-specific quirks without coupling to core release cycles. Thanks for the work!