# penfieldlabs/penfield-mcp [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/penfieldlabs/penfield-mcp  
**GitHub Stars:** 6  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/penfieldlabs-penfield-mcp

## Description
Penfield: persistent memory with hybrid search (BM25 + vector + graph), 24 relationship types for knowledge graphs, context checkpoints for cognitive handoff, artifact storage, and personality system. Works across Claude, Cursor, Windsurf, and any MCP client.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "penfield-mcp": {
    "command": "npx",
    "args": ["-y","mcp-remote"]
  }
}
```

## Documentation

## What penfieldlabs/penfield-mcp MCP server does

The penfieldlabs/penfield-mcp MCP server provides persistent memory for AI agents. It is designed to retain decisions, preferences, working context, and research findings beyond a single conversation or session. The service also represents connections between memories as a knowledge graph, allowing an agent to preserve relationships between ideas rather than storing isolated notes.

Its capabilities are organized into four groups:

- Memory tools for storing, recalling, searching, fetching, and updating memories.
- Knowledge graph tools for creating connections, removing connections, and exploring related information.
- Context tools for loading identity and personality context, reviewing recent work, saving checkpoints, restoring checkpoints, and listing saved contexts.
- Artifact tools for saving, retrieving, listing, and deleting stored artifacts.

The README describes 17 tools in total. Hybrid retrieval combines BM25 keyword matching, vector-based semantic search, and graph relationships. This lets an agent look for exact terms, related meaning, or connected memories depending on the task.

## How it works

Use the hosted MCP endpoint at `https://mcp.penfield.app` from a compatible client. After signing up through the Penfield portal, connect the endpoint and complete the OAuth authentication flow when prompted. The same memory and knowledge graph can then be used from different connected clients.

A recommended session pattern begins with `awaken()` to load identity and personality context, followed by `reflect()` to orient the agent on recent work. The default reflection period is the previous seven days. Without these calls, the README notes that an agent starts without prior context for the session.

The service supports Claude, Cursor, Windsurf, Gemini CLI, and other MCP-compatible tools. The repository also documents configurations for Cline, Roo Code, and Zed. An HTTP API is available separately at `api.penfield.app` for custom integrations, but that is distinct from the MCP connection described here.

## Setup and configuration

For Claude, add `https://mcp.penfield.app` as a custom connector. Claude Code can register it with an HTTP transport using the documented `claude mcp add` command. Cursor and other clients can use an MCP configuration that runs `npx -y mcp-remote https://mcp.penfield.app/`; the repository provides examples for Cursor, Windsurf, Cline, Roo Code, Zed, and Gemini CLI.

The penfieldlabs/penfield-mcp MCP server is hosted remotely, so local installation of the repository is not required by the documented setup. You need a Penfield account and OAuth authentication. The provided material does not state a pricing plan or quota.

## Tools and capabilities

The penfieldlabs/penfield-mcp MCP server exposes memory operations, graph relationships, context checkpoints, and artifact storage through MCP tools. It supports 11 documented memory types and 24 relationship types, with further parameter details in the repository's tool and reference documentation.

Typical uses include retaining architectural decisions across coding sessions, carrying investigation threads between sessions, building connected research notes, and checkpointing progress on long-running work. Artifact operations provide a separate way to store and retrieve files or other project outputs supported by the service.

## Limitations and notes

The material describes a hosted service and OAuth login but does not provide local deployment instructions, self-hosting details, or environment variables. It also does not specify pricing, retention limits, storage quotas, search-index behavior, or data residency. The native OpenClaw plugin mentioned in the README is a separate project and should not be confused with this MCP server.

_Full upstream README: https://allmcps.com/mcp/penfieldlabs-penfield-mcp/readme_