📚 Soulcraft Memory Documentation

Complete guide to using Soulcraft Memory - AI memory with 17 powerful tools via Model Context Protocol (MCP).

Quick Start 🤖 Agent Detection All 17 Tools Example Prompts Use Cases Setup Guides Troubleshooting

🚀 Quick Start

Get started with Soulcraft Memory in under 5 minutes:

  1. Sign in with Google or GitHub at memory.soulcraft.com/dashboard
  2. Copy your API key from the dashboard
  3. Choose your AI tool and follow the setup instructions
  4. Start using the 17 tools - just talk naturally to Claude!

🤖 Smart Agent Detection & Context Loading

Soulcraft Memory automatically detects your AI agent and optimizes for your workflow!

✨ What This Means for You

When you start a new session, Soulcraft Memory:

  • Detects your AI agent (Claude Code, Claude Desktop, Cursor, GitHub Copilot, VS Code, etc.)
  • Loads agent-appropriate context - Code agents get code context, chat agents get conversation history
  • Restores your session automatically - Active project, pending todos, preferences, recent work
  • Fixes the session isolation issue - No more memory loss on restart!

🎯 Supported AI Agents (27+ and Growing)

Tier 1 - Full Support & Testing:

  • ✅ Claude Code (CLI)
  • ✅ Claude Desktop
  • GitHub Copilot
  • VS Code
  • Cursor

Tier 2 - High Priority: Windsurf, Cline, JetBrains AI, Continue.dev, Sourcegraph Cody

Tier 3 - Community Supported: 10+ additional agents including Zed, Amazon Q, Tabnine, ChatGPT Desktop, Google Gemini, and more!

⚡ Context Loading Strategies

Each agent type gets optimized context:

  • Code-First (Claude Code, GitHub Copilot): Active project, recent code, decisions, errors (2000 tokens, 30 days)
  • Conversation-First (Claude Desktop, ChatGPT): Recent conversations, preferences, facts (1000 tokens, 7 days)
  • Workspace-First (Cursor, Windsurf): Workspace structure, multi-file context, todos (3000 tokens, 60 days)
  • Minimal (Terminal agents): Essential context only (500 tokens, 3 days)
  • Balanced (Default): Best of all strategies (1500 tokens, 14 days)

💡 Common Use Cases

🎯 Getting Started - Save Your Preferences

Tell Claude once, remember forever:

  • "Remember that I prefer TypeScript with 2-space indentation"
  • "I'm working on a project called 'brain-cloud'"
  • "Save this for later: The API endpoint is https://api.example.com"
  • "Note that Sarah is the project lead for the mobile app"

🔍 Finding Information

Search across all your memories:

  • "What do you remember about my coding preferences?"
  • "Search for anything about the database"
  • "What did we discuss about authentication yesterday?"
  • "Find my notes about the API design"

✅ Task Management

Never lose track of what needs to be done:

  • "Add a todo to implement user authentication"
  • "What are my pending todos for the brain-cloud project?"
  • "Mark the authentication todo as completed"
  • "Create a high-priority task to fix the login bug"

📂 Code Snippets & Files

Save reusable code and configurations:

  • "Save this auth helper function to /snippets/auth.ts"
  • "Show me the database config I saved earlier"
  • "Search for code snippets about API requests"

🛠️ All 17 MCP Tools

Soulcraft Memory provides 17 powerful tools organized into 5 categories. All tools work automatically when you talk naturally to Claude! 🆕 3 new tools for session continuity and personalization!

💬 Example Prompts for Each Tool

Below are detailed examples for all 17 tools. Just talk naturally - Claude will automatically use the right tool!

🧠 Memory Tools 6 tools

Store, search, and manage information that persists across all sessions.

memory_store

Store any information with automatic semantic indexing

Example Prompts:
  • "Remember that I prefer TypeScript with 2-space indentation"
  • "Save this for later: The production server is prod-server-01"
  • "Note that we decided to use PostgreSQL instead of MySQL"
  • "Keep in mind that the deadline for the mobile app is December 15th"
  • "Store this API key: abc123xyz456"

What happens: Claude automatically stores this in your memory with semantic tags, making it searchable and recallable across all sessions and devices.

memory_search

Search through memories using semantic similarity (finds by meaning, not just keywords)

Example Prompts:
  • "What do you know about my coding preferences?"
  • "Search for anything related to the database"
  • "Find information about authentication"
  • "What have I told you about the API design?"
  • "Show me memories about deadlines"

What happens: Searches semantically (by meaning), so "What's my coding style?" will find "I prefer TypeScript" even though the words don't match exactly.

memory_get

Retrieve a specific memory by ID

Example Prompts:
  • "Show me memory abc123"
  • "Retrieve the details of that preference I saved"
  • "Get memory #456"

What happens: Fetches the exact memory with all its metadata and relationships.

memory_update

Update existing memories with new information

Example Prompts:
  • "Update my indentation preference to 4 spaces"
  • "Change that server name to prod-server-02"
  • "Modify the deadline to January 30th"
  • "Update my TypeScript preference to include strict mode"

What happens: Finds and updates the relevant memory while preserving its history and relationships.

memory_delete

Remove memories you no longer need

Example Prompts:
  • "Delete that old API endpoint memory"
  • "Remove the note about the deadline"
  • "Forget about the PostgreSQL preference"

What happens: Permanently removes the memory and its relationships from your knowledge base.

memory_relate

Create semantic relationships between memories

Example Prompts:
  • "Connect my TypeScript preference to the brain-cloud project"
  • "Link the API endpoint to the authentication feature"
  • "Associate the deadline with the mobile app project"

What happens: Creates a relationship in the knowledge graph, enabling Claude to understand context and connections between different pieces of information.

✅ Todo Tools 3 tools

Create, manage, and track tasks that persist across all sessions.

todo_create

Create todos that persist across all sessions

Example Prompts:
  • "Add a todo to implement user authentication"
  • "Create a task: Write tests for the API endpoints"
  • "Remind me to update the documentation"
  • "Add a high-priority todo to fix the login bug"
  • "Create a task for the brain-cloud project: Deploy to production"

What happens: Creates a persistent todo with optional priority, project, and tags that you can track across all sessions.

todo_list

List todos with status, project, and priority filtering

Example Prompts:
  • "What are my pending todos?"
  • "Show me completed tasks"
  • "List all todos for the brain-cloud project"
  • "What high-priority tasks do I have?"
  • "Show me my todos from this week"

What happens: Returns all matching todos with their status, priority, and associated projects.

todo_update

Update todo status, priority, or details

Example Prompts:
  • "Mark the authentication todo as completed"
  • "Change the priority of the testing task to high"
  • "Update that todo to 'in progress'"
  • "Mark all documentation todos as done"

What happens: Updates the specified todo(s) with new status, priority, or details.

🎯 Project Context Tools 2 tools

Set and track your current project for automatic organization.

context_set

Set current project for automatic organization

Example Prompts:
  • "I'm now working on the brain-cloud project"
  • "Switch context to the mobile-app project"
  • "Set my current project to marketing-website"

What happens: Sets your active project context, so new memories and todos are automatically tagged with this project.

context_get

Get current project context and metadata

Example Prompts:
  • "What project am I currently working on?"
  • "Show me my current context"
  • "What's the active project?"

What happens: Returns your current project context with associated metadata and recent activity.

💬 Conversation Intelligence Tools 3 tools

Search, analyze, and learn from past conversations.

conversation_search

Search through past conversations semantically

Example Prompts:
  • "What did we discuss about the API yesterday?"
  • "Find conversations about authentication"
  • "Search for when we talked about database design"
  • "Show me our discussions about the mobile app"

What happens: Searches all past conversations semantically and returns relevant excerpts with context.

project_history

Get complete history for a project

Example Prompts:
  • "Show me the history for brain-cloud"
  • "What have we worked on together for the mobile app?"
  • "Give me a timeline of the marketing website project"

What happens: Returns a chronological history of all conversations, decisions, and changes for the specified project.

conversation_summary

Generate intelligent summaries of conversations

Example Prompts:
  • "Summarize our last conversation"
  • "Give me a recap of what we discussed today"
  • "What were the key points from our API discussion?"

What happens: Generates an intelligent summary highlighting key decisions, action items, and important points from the conversation.

🔄 Session Continuity & Personalization Tools 3 tools NEW!

Automatic session restoration and personalized AI experience. These tools work automatically!

session_resume

🎯 Automatically restores your context when you start a new session (Called automatically - no action needed!)

How It Works:
  • Called automatically when you start Claude - No prompts needed!
  • Detects your AI agent (Claude Code, Claude Desktop, Cursor, etc.)
  • Loads agent-appropriate context (active project, todos, preferences, recent work)
  • Uses smart token budgeting (500-3000 tokens based on agent type)
  • Filters by time (3-60 days back based on agent)
Manual Usage (Optional):
  • "Resume my previous session"
  • "Load my context"
  • "What was I working on last time?"

What happens: Loads your active project, pending todos, user preferences, recent decisions, and recent work. Context is optimized for your specific AI agent type!

✨ This fixes the #1 pain point: Session isolation! You never lose context when restarting your AI agent.

preferences_set

Store your preferences for a personalized AI experience

Example Prompts:
  • "Remember that I prefer TypeScript with 2-space indentation" (code_style)
  • "Save my preference: I always use React with TypeScript" (frameworks)
  • "Remember to always be concise in your responses" (agent_behavior)
  • "Set my privacy preference: never store sensitive data" (privacy)
  • "I prefer dark mode" (ui)
  • "Remember that I work 9am-5pm EST" (general)
Categories:
  • code_style: Indentation, naming conventions, formatting
  • frameworks: Preferred libraries, tools, tech stack
  • agent_behavior: How Claude should respond to you
  • privacy: What to remember/forget
  • ui: Interface preferences
  • general: Any other preferences
Scopes:
  • user: Global preference across all projects (default)
  • project: Specific to one project
  • agent: Specific to one AI agent (e.g., Claude Code vs Cursor)

What happens: Stores your preference in Brainy as NounType.State with proper categorization and scoping. Automatically loaded by session_resume!

preferences_get

Retrieve your saved preferences

Example Prompts:
  • "What are my coding preferences?"
  • "Show me my framework preferences"
  • "What preferences have I set?"
  • "Get my privacy settings"
  • "Show me project-specific preferences for brain-cloud"
Filters:
  • By category (code_style, frameworks, agent_behavior, privacy, ui, general)
  • By specific key
  • By scope (user, project, agent)
  • By scope ID (specific project or agent name)

What happens: Queries Brainy for preferences matching your filters and returns them in a structured format with category, key, value, scope, and last updated time.

⚙️ Setup Guides

Choose your AI coding tool below for specific setup instructions. Each tool has a different configuration approach.

🚀 Claude Code (Recommended: CLI Method)

Easiest setup! Claude Code is the only tool with native CLI commands for MCP configuration.

One-command setup:

claude mcp add --transport http soulcraft-memory \
  https://memory.soulcraft.com/mcp \
  -H "X-API-Key: YOUR_API_KEY_HERE" \
  -H "Accept: application/json"

Verify connection:

claude mcp list

Remove if needed:

claude mcp remove soulcraft-memory
Alternative: Manual Configuration (Advanced)

If you prefer to edit config files directly:

Config file location:

  • macOS/Linux: ~/.claude.json or ~/.config/claude/mcp.json
  • Windows: %USERPROFILE%\.claude.json or %APPDATA%\claude\mcp.json

Configuration:

{
  "mcpServers": {
    "soulcraft-memory": {
      "transport": "http",
      "url": "https://memory.soulcraft.com/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Restart Claude Code after saving.

🖥️ Claude Desktop (GUI App)

For the Claude Desktop application (not Claude Code CLI):

Config file location:

Configuration:

{
  "mcpServers": {
    "soulcraft-memory": {
      "command": "npx",
      "args": ["-y", "@soulcraft/mcp-server"],
      "env": {
        "SOULCRAFT_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Important: Completely quit and restart Claude Desktop after editing (not just refresh).

📱 Claude Mobile (iOS/Android)

Configure MCP servers in Claude Desktop or on claude.ai/settings, then your settings automatically sync to the mobile app!

🎯 Cursor

Config file location:

Configuration:

{
  "mcpServers": {
    "soulcraft-memory": {
      "transport": "http",
      "url": "https://memory.soulcraft.com/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Restart Cursor after saving.

🌊 Windsurf (Codeium)

Config file location:

Configuration:

{
  "mcp_servers": {
    "soulcraft-memory": {
      "command": "npx",
      "args": ["-y", "@soulcraft/mcp-server"],
      "env": {
        "SOULCRAFT_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Restart Windsurf after saving.

💻 GitHub Copilot (VS Code)

Option 1: VS Code Settings UI

  1. Open VS Code Settings (Cmd/Ctrl + ,)
  2. Search for "copilot mcp"
  3. Click "Edit in settings.json"
  4. Add the configuration below
  5. Restart VS Code

Option 2: Direct Edit

Add to .vscode/settings.json or ~/.config/Code/User/settings.json:

{
  "github.copilot.mcp.servers": {
    "soulcraft-memory": {
      "command": "npx",
      "args": ["-y", "@soulcraft/mcp-server"],
      "env": {
        "SOULCRAFT_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

🔧 Continue.dev

Config file location: .continue/mcpServers/soulcraft.yaml

Configuration:

name: Soulcraft Memory
command: npx
args:
  - "-y"
  - "@soulcraft/mcp-server"
env:
  SOULCRAFT_API_KEY: YOUR_API_KEY_HERE

Restart Continue.dev extension after saving.

🛠️ Other MCP-Compatible Tools

For tools not listed above, use this universal HTTP configuration:

{
  "mcpServers": {
    "soulcraft-memory": {
      "transport": "http",
      "url": "https://memory.soulcraft.com/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Check your tool's documentation for the config file location (usually ~/.your-tool/mcp.json).

🔧 Troubleshooting

Connection Issues

Can't Find Config File

Use this command to locate your config file:

find ~ -name "mcp.json" 2>/dev/null

Tools Not Showing Up

API Key Not Working

Still Having Issues?

We're here to help! Contact support at support@soulcraft.com or check our GitHub repository for community support.

← Back to Dashboard