Skip to content

Add OpenClaw skill integration#12

Closed
akadan18 wants to merge 1 commit intoMemPalace:mainfrom
akadan18:integrations/openclaw
Closed

Add OpenClaw skill integration#12
akadan18 wants to merge 1 commit intoMemPalace:mainfrom
akadan18:integrations/openclaw

Conversation

@akadan18
Copy link
Copy Markdown

@akadan18 akadan18 commented Apr 7, 2026

What this adds

An OpenClaw skill file (integrations/openclaw/SKILL.md) that teaches OpenClaw agents how to use MemPalace.

OpenClaw is an open-source personal AI harness — the skill file is how agents learn to use tools. With this file, anyone using OpenClaw can install and use MemPalace with a single command:

clawhub install mempalace

What's in the skill file

  • Setup and install instructions
  • CLI command reference (search, mine, wake-up, AAAK)
  • MCP server setup for Claude/GPT
  • Mining modes explained

Why

MemPalace's MCP integration already works great with Claude Code. This extends that to OpenClaw harnesses where agents can call mempalace_search automatically during sessions.

Built by GraceClaw — a production multi-agent setup running on OpenClaw.

@vincentkoc
Copy link
Copy Markdown

OpenClaw maintainer here, native plugin support would be better as we have memory hooks.

igorls added a commit to igorls/mempalace that referenced this pull request Apr 7, 2026
- Remove palace_path from _no_palace() error response (prevents
  leaking filesystem paths to the LLM)
- Replace str(e) with generic 'Internal tool error' in MCP dispatch
  catch block (full error is still logged server-side via stderr)
- Replace sys.exit(1) with return in searcher.search() CLI function
  (prevents process termination if called from library context)
- Remove unused sys import from searcher.py

Findings: MemPalace#12 (HIGH), MemPalace#5 (MEDIUM), MemPalace#15 (LOW)

Includes test infrastructure from PR MemPalace#131.
92 tests pass.
@bensig
Copy link
Copy Markdown
Collaborator

bensig commented Apr 7, 2026

Thanks for the OpenClaw integration! This is pretty niche though — we'd rather keep integrations in their own repos for now. Feel free to publish it as a standalone package and we can link to it.

@bensig bensig closed this Apr 7, 2026
igorls added a commit to igorls/mempalace that referenced this pull request Apr 7, 2026
- Remove palace_path from _no_palace() error response (prevents
  leaking filesystem paths to the LLM)
- Replace str(e) with generic 'Internal tool error' in MCP dispatch
  catch block (full error is still logged server-side via stderr)
- Replace sys.exit(1) with return in searcher.search() CLI function
  (prevents process termination if called from library context)
- Remove unused sys import from searcher.py

Findings: MemPalace#12 (HIGH), MemPalace#5 (MEDIUM), MemPalace#15 (LOW)

Includes test infrastructure from PR MemPalace#131.
92 tests pass.
jphein referenced this pull request in jphein/mempalace Apr 27, 2026
…dential

Restores daemon-routing to the committed plugin manifest so JP's fork
main carries the daemon-routed config matching production deployment,
but **without** re-embedding the literal API key.

The previous shape had the key inline:

```json
"env": {"PALACE_API_KEY": "adb43..."}
```

The new shape drops the env block entirely. PALACE_API_KEY is
inherited at runtime from `~/.claude/settings.local.json`'s env block
(which Claude Code passes through to spawned MCP servers and hooks).
That file is gitignored / 0600-perm / per-machine, so the key lives
exactly one place: in JP's local settings, not in any committed source.

Daemon URL and the mempalace-mcp.py client path remain literal —
they're machine-specific identifiers (homelab address + JP's
palace-daemon checkout path), not credentials. They identify *what*
to talk to, not *how to authenticate*.

Anyone forking this repo would rewrite both for their own setup; the
committed values are JP-development-fork defaults, not portable
defaults. The PR-ready shape is what upstream MemPalace's manifest
already uses (`{"command": "mempalace-mcp"}`) — that ships via
contributions to upstream, not via this fork main.

hooks/hooks.json is unchanged (already at upstream-shape after the
earlier scrub) — Claude Code passes the env block to bash hooks too,
so no prelude needed.

Companion to upstream PR #12 (which fixes the same class of
embedded-default in palace-daemon's `clients/palace-mode`).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants