Memex AI Connectors

Use Anthropic, OpenAI, and Gemini models powered by Memex Credits—no external API accounts required.

What Are Memex AI Connectors?

Memex AI Connectors let you add AI capabilities to your applications using your Memex Credits. Instead of requiring API keys from Anthropic, OpenAI, or Google, your app's AI usage is charged against your Memex Credits.

These connectors are designed for applications where you want to provide AI features seamlessly.

Three providers are available:

Connector	Provider
Anthropic	Claude models
OpenAI	GPT models
Gemini	Gemini models

Already have your own API key? If you're building apps for yourself and have your own provider account, use the regular connectors instead: Anthropic, OpenAI, or Gemini. Those connectors bill your provider account directly.

Enabling a Connector

1. Open the Memex Hub from the sidebar

2. Click on the Connectors tab

3. Click on the connector card for your chosen provider (Anthropic, OpenAI, or Gemini)

4. Click Create Connection

Memex handles the rest automatically—generating a secure API key, testing the connection, and configuring everything behind the scenes.

Your connector is now active and available in all your projects. You can enable multiple connectors (even all three).

To remove a connector, open its card and click Delete Connection. This automatically revokes the associated key and removes the connector from all projects.

Building Apps with Your Connection

Reference your connector in a conversation using @ to link it, then describe what you want to build:

@MY_CONNECTOR Build a document Q&A app where users can upload PDFs and ask questions about the content.

Choosing a Provider

All three connectors work the same way—the main differences are the models available and the capabilities each provider offers. You can add multiple connectors and switch between them.

Switching providers: Each provider has a different SDK and API format, so changing from one to another may require the agent to adapt your app's code. If you want to switch, ask the agent: "Switch this app from OpenAI to Anthropic" and it will handle the migration.

Anthropic (Claude)

Claude models excel at careful reasoning, long-form writing, and following nuanced instructions. Ask the agent to use a specific model by name, e.g. "Use Sonnet 4.5 for this app".

Model	Best For
Opus 4.6	Most capable, complex tasks
Sonnet 4.5	Balanced performance and cost

For the full model list, see the Anthropic Models Documentation.

Extra capabilities you can ask the agent to use:

• Extended thinking for complex reasoning tasks

• PDF and image analysis (vision)

• Structured JSON output

OpenAI (GPT)

GPT models offer broad general-purpose capabilities including text, image generation (DALL-E), and function calling. Ask the agent to use a specific model, e.g. "Use gpt-4.1 for this app".

Model	Best For
gpt-4.1	Most capable general-purpose model
gpt-4.1-mini	Fast, cost-effective responses
gpt-4.1-nano	Fastest, lowest cost

For the full model list, see the OpenAI Models Documentation.

Extra capabilities you can ask the agent to use:

• Image generation with DALL-E

• JSON mode for structured output

• Function calling for tool use

Gemini

Gemini models are strong at multimodal tasks, long context windows, and grounded responses. Ask the agent to use a specific model, e.g. "Use gemini-2.0-flash for this app".

Model	Best For
gemini-2.0-flash	Fast, balanced performance
gemini-2.5-pro-preview-06-05	Most capable, complex tasks

For the full model list, see the Google Gemini API Documentation.

Extra capabilities you can ask the agent to use:

• Long context (up to 1M tokens)

• Multimodal input (images, audio, video)

• Grounded responses with Google Search

Credit Usage

Each API call consumes Memex Credits based on the model used and the number of tokens processed. Monitor your credit usage in Settings → Usage and Billing.

For more details on credits, see Plans and Billing.

Security

Memex handles all API key management automatically:

• Keys are generated securely on your behalf

• Keys are stored encrypted (SHA-256 hashed server-side)

• Keys are automatically revoked when you delete a connector

• You don't need to manage or rotate keys manually

Troubleshooting

You can ask the Memex agent to help troubleshoot, but here are a few common problems:

"Connection Failed" Error

• Verify your internet connection

• Try disabling and re-enabling the connector

"Insufficient Credits" Error

• Check your credit balance in Settings → Usage and Billing

• Purchase additional credits or upgrade your plan

Learn More

• Anthropic Models Documentation

• OpenAI API Documentation

• Google Gemini API Documentation

• Memex Credits and Billing