← Back to Skills Marketplace
245
Downloads
2
Stars
0
Active Installs
4
Versions
Install in OpenClaw
/install pieces-mcp
Description
Connect OpenClaw to Pieces via MCP-only (no SSE) and use Pieces as external long-term memory. Use this when the human runs PiecesOS with LTM enabled on anoth...
Usage Guidance
This skill appears coherent for connecting OpenClaw to a Pieces MCP server via a tunnel. Before installing or using it: 1) Be aware it will instruct you to expose a PiecesOS MCP endpoint to the public internet via ngrok or a custom tunnel — only do this if you trust the remote machine and understand the exposure risk. Use short-lived tunnels and restrict access where possible. 2) Do not paste or share any secret tokens, SSH keys, or private URLs in public chat; the skill only needs the HTTPS tunnel URL and the server-assigned session id pattern. 3) Confirm any instructions that tell you to install or run third-party tooling (mcp-remote, MCPorter) come from trusted sources before you run them. 4) If additional parts of the SKILL.md (not included here) instruct running downloads from arbitrary URLs or ask for unrelated credentials, treat that as suspicious and stop. If you want a stricter assessment, provide the rest of SKILL.md or any lines that instruct installing binaries or running remote scripts.
Capability Analysis
Type: OpenClaw Skill
Name: pieces-mcp
Version: 1.0.3
The skill bundle provides detailed instructions for an OpenClaw agent to integrate with Pieces Long-Term Memory using the Model Context Protocol (MCP). It guides the agent through setting up a secure tunnel (e.g., ngrok), installing the 'mcp-remote' npm package, and configuring 'mcporter.json'. The instructions include security-conscious practices, such as using file-based JSON for curl commands to prevent shell injection, and the overall logic is transparently aligned with the stated purpose of external memory integration.
Capability Assessment
Purpose & Capability
The name/description describe integrating OpenClaw with Pieces MCP and the instructions focus on constructing and using an MCP URL, session handling, and using Pieces tools. The SKILL.md references MCPorter and mcp-remote (expected bridge components) but does not require unexpected credentials or unrelated system access.
Instruction Scope
Instructions are narrowly about validating a user-provided HTTPS tunnel, initializing MCP sessions, and making properly formed requests. The skill asks the agent to prompt the human to run ngrok or a similar tunnel and to paste the tunnel URL. This requires the agent to instruct the human to expose a local service to the internet (a legitimate but security-sensitive operation). The SKILL.md does not appear to instruct the agent to read unrelated host files or exfiltrate secrets.
Install Mechanism
This is an instruction-only skill with no install spec and no code files — lowest-risk install surface. It references external tools but does not attempt to download or execute code itself.
Credentials
The skill declares no required environment variables, secrets, or config paths. The runtime instructions rely on a user-supplied HTTPS URL and server-assigned session IDs only, which is proportionate to the stated purpose.
Persistence & Privilege
always is false and the skill does not request permanent or system-wide modifications. Nothing in the SKILL.md requests modifying other skills or global agent settings.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install pieces-mcp - After installation, invoke the skill by name or use
/pieces-mcp - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.3
No functional changes detected; documentation update only.
v1.0.2
- Emphasizes file-based JSON for curl and string IDs for JSON-RPC requests to prevent shell/formatting errors.
- Clarifies session management: always extract and use the server-assigned session ID from the response header.
- Enhances troubleshooting and error-handling instructions for connecting to the MCP server.
- Streamlines setup steps and improves language for guiding human users.
v1.0.1
Added exact example requests for initializing Pieces, asking Pieces, and creating memories in Pieces.
v1.0.0
pieces-mcp v1.0.0
- Initial release of the Pieces MCP Skill for OpenClaw.
- Enables OpenClaw agents to connect to Pieces as long-term memory using MCP-only (no SSE) via a remote MCP server exposed over an ngrok tunnel.
- Provides step-by-step instructions for setup, including human guidance, MCPorter configuration, and connection verification.
- Documents correct usage of Pieces tools (`ask_pieces_ltm`, `create_pieces_memory`, search functions) as part of agent reasoning.
- Includes troubleshooting guidelines for common setup and connectivity issues.
Metadata
Frequently Asked Questions
What is Pieces Long-Term Memory (MCP)?
Connect OpenClaw to Pieces via MCP-only (no SSE) and use Pieces as external long-term memory. Use this when the human runs PiecesOS with LTM enabled on anoth... It is an AI Agent Skill for Claude Code / OpenClaw, with 245 downloads so far.
How do I install Pieces Long-Term Memory (MCP)?
Run "/install pieces-mcp" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Pieces Long-Term Memory (MCP) free?
Yes, Pieces Long-Term Memory (MCP) is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Pieces Long-Term Memory (MCP) support?
Pieces Long-Term Memory (MCP) is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Pieces Long-Term Memory (MCP)?
It is built and maintained by Jack Ross (@jackrosspieces); the current version is v1.0.3.
More Skills