ChatGPT

Connect ChatGPT to Basic Memory using remote MCP integration

ChatGPT connects to Basic Memory through OpenAI’s Model Context Protocol implementation, providing search and document retrieval from your knowledge base.

Requirements

ChatGPT Plus or Pro subscription required

MCP integration is not available on ChatGPT’s free tier.

  • ChatGPT Plus: $20/month
  • ChatGPT Pro: $200/month (includes advanced features)

Remote MCP server required

ChatGPT only supports remote MCP servers, not local connections like Claude Desktop.

Available Tools

By default, ChatGPT uses two tools: search for finding content and fetch for retrieving documents.

For full MCP access, explicitly select and enable the Basic Memory MCP server in ChatGPT’s developer mode. Once enabled, ChatGPT can use all 17+ MCP tools including write_note, read_note, edit_note, and project management tools.

Search across your knowledge base.

Parameters:

  • query (string, required) - Search query

Example:

You: "Search my knowledge base for authentication notes"

ChatGPT internally calls:
{
  "tool": "search",
  "arguments": {
    "query": "authentication notes"
  }
}

Response:
I found 2 documents about authentication:
1. Auth Design
2. OAuth Setup

Search response includes:

  • Document ID (permalink)
  • Document title
  • Document URL
  • Total count of results
  • Original query

Results limited to 10 per search.

fetch

Retrieve the full contents of a document.

Parameters:

  • id (string, required) - Document identifier (permalink)

Example:

You: "Show me the Auth Design document"

ChatGPT internally calls:
{
  "tool": "fetch",
  "arguments": {
    "id": "auth-design"
  }
}

Response:
Here's the Auth Design document:

# Auth Design

## Overview
...full content...

Fetch response includes:

  • Document ID
  • Document title
  • Full markdown content
  • Document URL
  • Metadata (format type)

Usage Examples

Search Workflow

You: "Find my notes about database design"
ChatGPT: [Calls search tool with query "database design"]
ChatGPT: I found 3 relevant documents:
         1. Database Schema Design
         2. PostgreSQL Performance Tips
         3. Database Migration Strategy

Fetch Workflow

You: "Show me the Database Schema Design document"
ChatGPT: [Calls fetch tool with id "database-schema-design"]
ChatGPT: [Displays full document content]

Combined Search and Fetch

You: "Find documentation about API endpoints and show me the details"
ChatGPT: [Searches for "API endpoints"]
ChatGPT: I found API Documentation. Let me fetch the details.
ChatGPT: [Fetches and displays full API documentation]

Configuration

Project Resolution

ChatGPT tools automatically resolve projects using:

  1. CLI --project flag constraint (if MCP server started with this flag)
  2. Default project mode configuration
  3. Error if no project can be resolved

Recommended configuration:

{
  "default_project": "main",
  "default_project_mode": true
}

Limitations

ChatGPT-Specific Constraints

  1. Remote only - No local MCP server support
  2. Two tools by default - search and fetch (17+ tools available when MCP explicitly enabled)
  3. Fixed pagination - 10 results per search, not configurable
  4. Simplified parameters - No advanced filtering options for default tools
  5. Subscription required - ChatGPT Plus or Pro only
  6. Network latency - Remote calls slower than local operations

Comparison with Claude Desktop

FeatureChatGPTClaude Desktop
MCP ModeRemote onlyLocal or Remote
Tools2 by default, 17+ when enabled17+ (full suite)
Project SupportFull multi-projectFull multi-project
SubscriptionPlus/Pro requiredFree (Claude)
ConfigurationDeveloper modeConfig file
PerformanceNetwork latencyLocal (instant)

Next Steps

Built with ❤️ by Basic Memory