Pensieve

Multi-user MCP server for saving, loading, listing, and searching conversation histories with JWT auth and cloud deployment.

html

0

GitHub Stars

html

Language

7 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "emjayahn-pensieve-mcp": {
      "command": "uv",
      "args": [
        "run",
        "python",
        "-m",
        "mcp_server.server"
      ]
    }
  }
}

Pensieve MCP Server enables multi-user, cloud-hosted sharing of conversation history between ChatGPT and Claude, with secure storage, search, and easy management of conversations.

How to use

Connect Claude to the Pensieve MCP Server by configuring a local MCP client profile that points to the Pensieve MCP endpoint. You can run the MCP server locally or deploy it to the cloud, then use the provided tools to save, load, search, and manage conversations. Each user has an isolated space, and conversations can be appended with new messages.

Save a conversation to Pensieve to capture the chat history and optional metadata like title or tags. Use the save_conversation tool from your MCP client to persist the current session.

Load a previously saved conversation by its ID to review or continue an earlier chat. Use the load_conversation tool with the specific conversation ID.

Search for conversations by keywords to locate relevant history quickly. Use the search_conversations tool and provide search terms.

List all saved conversations to get an overview of available history. This helps you identify IDs for loading or archiving.

How to install

Prerequisites: you need Python and a functional environment for running the MCP server. You also need uv to execute Python modules directly.

# Step 1: Clone the Pensieve MCP repository
git clone <repository-url>
cd pensieve-mcp

# Step 2: Install dependencies
uv pip install -e .

Configuration and deployment notes

You can run the MCP server locally or deploy to Azure Container Apps for cloud hosting. The server exposes a standard interface for MCP clients like Claude and supports multi-user authentication via JWT tokens.

Local startup example (stdio configuration for an MCP client): use uv to run the Python module that serves the MCP API.

{
  "mcpServers": {
    "pensieve-mcp": {
      "command": "uv",
      "args": ["run", "python", "-m", "mcp_server.server"],
      "cwd": "/path/to/pensieve-mcp"
    }
  }
}

Azure deployment and cloud notes

If you want cloud hosting, Pensieve can be deployed to Azure Container Apps, with a FastAPI backend and MongoDB-compatible Cosmos DB for storage, alongside JWT-based authentication.

Deployment steps include preparing Azure prerequisites, deploying the container, and configuring the MCP client to point at the cloud API URL.

Configure the MCP client API URL in your environment to enable cloud usage.

Available tools

save_conversation

Saves the current conversation to Pensieve with optional metadata like title and tags.

load_conversation

Retrieves a saved conversation by its ID for review or continuation.

search_conversations

Searches saved conversations by keywords to locate relevant history quickly.

list_conversations

Lists all saved conversations to provide an overview and IDs for loading or management.

append_to_conversation

Adds new messages to an existing conversation to extend history without creating a new entry.