Spark
Sign In
MCP Connector

Manage AI Agent Memory Spaces

An MCP server for Stitch AI's decentralized memory hub - create/delete named spaces and upload, retrieve, or list AI agent memories via 6 tools.

Connect
StitchAI
Maintainer?
90
Spark score
out of 100
Updated Apr 2025
Fresher alternatives ↓
Version 1.0.0
Models
universal
Add to Favorites

Why it matters

Integrate Stitch AI's memory management system into your AI agent infrastructure. This asset provides tools to create, retrieve, and manage AI agent memory within decentralized knowledge spaces.

Outcomes

What it gets done

01

Create and delete memory spaces.

02

Upload, retrieve, and manage memory content.

03

View all available memory spaces with filtering and pagination.

Install

Add it to your toolbox

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/vb-stitch-ai | bash

Capabilities

Tools your agent gets

create_space

Creates a new memory space with the specified name and type

delete_space

Deletes a memory space with the specified name

get_all_spaces

Retrieves a list of all available memory spaces

upload_memory

Uploads new memory to the specified memory space

get_memory

Retrieves specific memory by ID from a memory space

get_all_memories

Retrieves all memory from the specified memory space with optional filtering and pagination

Overview

Stitch AI MCP Server

An MCP server exposing six tools to create and delete named memory spaces and to upload, retrieve, or list AI agent memories within them, for Stitch AI's memory management system. Use when an AI agent needs a dedicated, space-organized memory store rather than folding memories into a general-purpose database or the conversation context.

What it does

Implements a Model Context Protocol server for Stitch AI's memory management system, described as a decentralized knowledge hub for AI, exposing six tools for creating, retrieving, and managing AI agent memories. create_space and delete_space manage named memory spaces (create takes a space name and a type; delete takes just the name), get_all_spaces lists every available space with no parameters, upload_memory writes a new memory into a named space (a space name, a message, and the memory content itself), get_memory retrieves one specific memory by ID from a space, and get_all_memories retrieves every memory from a space with optional filtering by comma-separated memory names, a result limit defaulting to 50, and an offset defaulting to 0 for pagination.

When to use - and when NOT to

Use it when an AI agent needs a dedicated, space-organized store for its own memories rather than folding them into a general-purpose database or the conversation context itself. Running the server locally is npm run start; wiring it into Claude Desktop means cloning the repository, installing the MCP SDK and zod as dependencies, and editing the platform-specific Claude Desktop config file - ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, or the AppData equivalent on Windows - to point at the cloned server entry point with an API key and base URL, then restarting Claude Desktop for the MCP tools to appear.

Capabilities

Six tools covering the full memory lifecycle: creating and deleting named memory spaces, listing all spaces, uploading a memory into a space, and retrieving either one memory by ID or all memories in a space with optional name filtering and pagination.

How to install

Clone the repository, install @modelcontextprotocol/sdk and zod plus TypeScript dev dependencies, then register the server in Claude Desktop's config pointing at the cloned entry point:

{
    "mcpServers": {
        "stitchai": {
            "command": "npx",
            "args": [
                "ts-node",
                "/path/to/cloned/stitch-ai-mcp/src/server.ts"
            ],
            "env": {
                "API_KEY": "<STITCH_AI_API_KEY>",
                "BASE_URL": "https://api-demo.stitch-ai.co"
            }
        }
    }
}

Who it's for

AI agent builders who want a dedicated, queryable memory store organized into named spaces, accessible to Claude Desktop or any other MCP client, instead of building memory persistence from scratch. The project is early-stage (versioned 0.1.0), positioned as a decentralized alternative to folding agent memory into a conventional centralized database, with support and updates tracked through the Stitch AI team's own X account rather than a dedicated support channel, and no separate testing or CI documentation beyond the basic build and run commands.

Source README

image
StitchAI
version

Stitch AI's MCP Server

Decentralized Knowledge Hub for AI

This repository contains a Model Context Protocol (MCP) server implementation for Stitch AI's memory management system. The server provides tools for creating, retrieving, and managing AI agent memories.

Available Tools

The MCP server provides the following tools:

create_space

Creates a new memory space with the specified name.

  • Parameters:
    • space_name: The name of the memory space to create
    • type: The type of memory space to create

delete_space

Deletes a memory space with the specified name.

  • Parameters:
    • space_name: The name of the memory space to delete

get_all_spaces

Gets a list of all available memory spaces.

  • Parameters: None

upload_memory

Uploads a new memory to a specified memory space.

  • Parameters:
    • space: The name of the memory space to upload to
    • message: The memory message to upload
    • memory: The memory content to upload

get_memory

Retrieves a specific memory by ID from a memory space.

  • Parameters:
    • space: The name of the memory space
    • memory_id: The ID of the memory to retrieve

get_all_memories

Retrieves all memories from a specified memory space.

  • Parameters:
    • space: The name of the memory space to retrieve memories from
    • Optional Parameters:
      • memory_names: Comma-separated list of memory names to filter
      • limit: Maximum number of memories to return (default: 50)
      • offset: Number of memories to skip (default: 0)

Run the server

npm run start

Using with Claude Desktop

  1. Clone the repository

    git clone https://github.com/StitchAI/stitch-ai-mcp.git

  2. Install dependencies

    npm install @modelcontextprotocol/sdk zod
npm install -D @types/node typescript

  3. Install Claude for Desktop

    • Download and install the latest version from Claude's website

  4. Configure Claude for Desktop

    • Locate your Claude for Desktop configuration file:
      • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
      • Windows: %AppData%\Claude\claude_desktop_config.json
    • Create the file if it doesn't exist

  5. Edit Configuration File

    • Open the configuration file in a text editor:
      • macOS: code ~/Library/Application\ Support/Claude/claude_desktop_config.json
      • Windows: code $env:AppData\Claude\claude_desktop_config.json
    • Add your MCP server configuration:
{
    "mcpServers": {
        "stitchai": {
            "command": "npx",
            "args": [
                "ts-node",
                "/path/to/cloned/stitch-ai-mcp/src/server.ts"
            ],
            "env": {
                "API_KEY": "<STITCH_AI_API_KEY>",
                "BASE_URL": "https://api-demo.stitch-ai.co"
            }
        }
    }
}
  1. Restart Claude for Desktop
    • After saving the configuration file, restart Claude for Desktop
    • The MCP UI elements will appear in Claude for Desktop once at least one server is properly configured

FAQ

Common questions

Discussion

Questions & comments · 0

Sign In Sign in to leave a comment.