Skip to content

MCP server for AI memory - store and retrieve context across AI conversations

License

Notifications You must be signed in to change notification settings

Contextable-me/mcp

Repository files navigation

@contextable/mcp

Give your AI a memory. Works with Claude Desktop and any MCP-compatible client.

Quick Start (30 seconds)

Add to your Claude Desktop config:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "contextable": {
      "command": "npx",
      "args": ["@contextable/mcp"]
    }
  }
}

Restart Claude Desktop. Done.

Now ask Claude: "Create a project called My App to track my development decisions"

What You Can Do

Save context that persists across conversations:

  • "Save this architecture decision to the project"
  • "Remember this API design for later"
  • "Store this bug investigation so I don't forget"

Find anything instantly:

  • "Search for authentication"
  • "What did we decide about the database?"
  • "Load all security-related artifacts"

Pick up where you left off:

  • "Resume my project"
  • "Show me the project summary"
  • "What were we working on?"

Features

Feature Description
Projects Organize context by project or topic
Artifacts Save decisions, code snippets, docs, conversations
Full-Text Search Find anything across all projects
Version History Every change tracked, rollback anytime
Auto-Chunking Large content automatically split
Topic Clustering Auto-detect themes across artifacts

All 13 Tools

Projects

  • project_save - Create or update a project
  • project_list - List all projects
  • project_resume - Load project with summaries
  • project_analysis_get - Get AI-generated insights

Artifacts

  • artifact_save - Save content (auto-chunks large files)
  • artifact_list - List with size estimates
  • artifact_get - Load full content
  • artifact_delete - Archive (recoverable)
  • artifact_restore - Restore archived
  • artifact_archived - List archived items
  • artifact_versions - View history
  • artifact_rollback - Restore previous version

Search

  • search - Full-text search across everything

Your Data

Everything is stored locally in SQLite:

~/.contextable/data.db

No account. No cloud. No tracking. Your data stays on your machine.


Want More?

Contextable Cloud adds:

Feature Local Cloud
Claude Desktop
ChatGPT -
Claude.ai (web) -
Sync across devices -
AI analysis & insights -
Team sharing - Coming soon

Try Contextable Cloud →


Configuration

Environment Variables

Variable Default Description
CONTEXTABLE_DATA_DIR ~/.contextable Data directory
CONTEXTABLE_DB_PATH ~/.contextable/data.db SQLite database path
CONTEXTABLE_LOG_LEVEL info Logging: debug, info, warn, error

Custom Database Location

{
  "mcpServers": {
    "contextable": {
      "command": "npx",
      "args": ["@contextable/mcp"],
      "env": {
        "CONTEXTABLE_DB_PATH": "/path/to/my/data.db"
      }
    }
  }
}

HTTP Mode (Advanced)

For web-based MCP clients, run as HTTP server:

npx @contextable/mcp --sse --port 3000

Endpoints:

  • GET /health - Health check
  • POST /mcp - MCP messages (JSON-RPC)

Programmatic Usage

import { SQLiteAdapter } from '@contextable/mcp';

const storage = new SQLiteAdapter({ path: './my-context.db' });
await storage.initialize();

// Create a project
const project = await storage.createProject({
  name: 'My Project',
  description: 'Project description',
});

// Save an artifact
const artifact = await storage.createArtifact({
  projectId: project.id,
  title: 'Design Decision',
  artifactType: 'decision',
  content: '# We chose PostgreSQL because...',
});

// Search
const results = await storage.search('postgresql');

await storage.close();

Development

git clone https://github.com/Contextable-me/mcp.git
cd mcp
npm install
npm run build
npm test

License

Apache 2.0 - See LICENSE

Links

About

MCP server for AI memory - store and retrieve context across AI conversations

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •