The platform works with any client that supports the Model Context Protocol over HTTP. You can connect Claude, Claude Code, ChatGPT, Codex, Cursor, Windsurf, GitHub Copilot, Gemini, Grok, Mistral, Perplexity and LM Studio without writing custom integration code.

Before you begin, you need a Giant Context account with at least one active project. Once your account is ready, you can choose the authentication method that best fits your AI client, your security needs and your daily workflow.

OAuth is the recommended authentication method for end users on Claude, ChatGPT, Cursor and most standard clients. There are no keys to copy and no manual credential step.

First, open your AI client, add a custom connector and enter the server URL https://mcp.giantcontext.com/mcp. Second, authorize the connection by signing in to your Giant Context account and granting access when prompted in your browser.

Finally, return to your client and call the welcomeMcp tool. This initial call returns a complete platform guide to your AI. It explains how to navigate your project before you describe what you want to build in your workspace.

If you are a developer using Claude Code, custom scripts or a client that requires manual MCP configuration, you should use an API key. This method bypasses the browser sign-in flow, making it ideal for programmatic environments and automated workflows.

To generate your credentials, open the Giant Context console and navigate to the API keys section to create a new key. You will only see this key once, so store it securely in your environment variables.

Next, configure your developer tool to send this API key as a bearer token with every request to the hosted server.

Once connected, you can manage your project in plain language. For example, tell the AI to build a pricing page for your product with three tiers. The AI generates a draft pricing page and returns it for review with a visual preview and a quality score.

You can also ask it to write a welcome email for new signups and send it when someone joins the early-access list. Alternatively, tell it to add a contact form to your homepage and route submissions to your CRM.

Beyond these examples, you can generate website posts, maintain a knowledge base, manage CRM contacts, handle files and run the Mind autonomy loop. The AI does the work, and you review it.

You never have to worry about the AI breaking your live site. By default, nothing goes live without your approval. Every change the AI proposes arrives as a safe, reviewable draft rather than an immediate live update.

Each draft includes a visual preview and a quality score to help you evaluate the proposed changes before it goes live. You can approve the draft to publish it, reject it entirely or ask the AI to revise it from the conversation.

This safety net applies to all operations, including autonomous actions triggered by the Mind loop. Every modification carries a full audit trail. You always maintain complete control over your project and its public-facing content.

When you connect an AI agent to Giant Context, it can never do more than the person who authorized it.

The AI cannot bypass your organization's security policies. Each team member connects their own account. There is no shared access. Every action the AI takes appears in the activity log under your name. This keeps the team accountable.

You can revoke access at any time. Disconnect the connector in your AI client or revoke the API key directly in the console. This immediately severs the connection and halts any further AI operations.

If you prefer a programmatic path instead of a chat client, Giant Context offers a full REST API alongside TypeScript and Python SDKs. You can explore these tools and their capabilities by visiting the /developers documentation portal.

To get started with any of these connection methods, you need an active account. Visit the /early-access page to sign up and create your first project. For complete details on data collection, security practices and usage, review the platform privacy policy at /privacy.