Overview of Top MCP's You Should be Using for Claude, ChatGPT and Gemini
This episode is a practical introduction to MCPs (Model Context Protocols) and how they can dramatically expand what AI assistants like Claude, ChatGPT, and Gemini can do. The host explains MCPs as a standardized way for AI tools to discover and use external software, files, and services, then walks through real-world examples using AI Box to generate images, audio, video, and text through connected workflows.
The main message: MCPs make AI agents more powerful, more reliable, and easier to connect to the tools you already use—especially for repetitive, multi-step, or high-volume workflows.
MCPs vs. APIs: The Core Idea
What an API is
- An API is the technical interface that lets one piece of software talk to another.
- It is often strict, structured, and specific to each product.
- If the call is formatted incorrectly, it can fail or crash.
What an MCP is
- An MCP is presented as a more universal, AI-friendly layer that helps an assistant understand how to use a tool or service.
- Instead of forcing the AI to infer everything from a long documentation page, the MCP gives it a more standardized way to interact with the software.
- The host compares it to a universal adapter: one that makes different tools easier for AI to use.
Why this matters
- With APIs, the AI often has to read documentation, guess structures, and retry many times.
- With MCPs, the assistant can connect more directly and reliably to the service.
Why MCPs Are Useful in Practice
The episode emphasizes three major benefits:
- Less friction: You don’t have to manually wrestle with API docs and formatting.
- More reliability: Fewer crashes and fewer failed integrations.
- More capability: You can give tools to Claude or ChatGPT that they do not natively have, like generating images, audio, or video.
The host also points out that MCPs can save tokens and reduce glitches compared with browser/computer-use workflows.
Real-World Examples from AI Box
1. AI-generated images, video, and audio
The host describes building an MCP for AI Box that lets:
- Claude generate images
- ChatGPT generate video
- Gemini generate audio through ElevenLabs
This effectively allows the main AI assistants to use capabilities they don’t inherently provide on their own.
2. Podcast publishing workflows
The host used podcast distribution platforms like:
- RSS.com
- Acast
- ART19
He explains how painful API-based integrations can be when every service has different rules, docs, and error handling. MCPs simplify that workflow by making it easier for AI agents to interact with the publishing systems.
3. Large-scale Bible verse generation
A major example involved generating explanations for every verse in the Bible:
- Tens of thousands of outputs
- High quality required
- Long-running process prone to failure
The host explains that AI Box helped by:
- Running generation in parallel
- Preserving outputs even if Claude or the connection crashes
- Preventing loss of work on large batch jobs
4. High-cost audio generation with ElevenLabs
He shares a painful story where a generation workflow crashed after spending about $1,200 in credits with no output saved. The point: if the generation engine is running through AI Box with MCP support, the assets remain safe even if the AI assistant fails mid-process.
Top MCPs and Connectors Mentioned
Most useful for builders
The host recommends these as a strong default stack:
- GitHub MCP
- Vercel MCP or Netlify MCP
- Supabase MCP
Together, these let Claude help with:
- Code storage and versioning
- Hosting and deployment
- Database setup and management
Common productivity connectors
He also highlights:
- Gmail
- Google Calendar
- Google Drive
These are useful for everyday work and knowledge management.
Other popular MCPs
Some of the most useful or commonly recommended ones include:
- Brave Search — for current web search, current releases, CVEs, and up-to-date info
- Figma — for design-to-code workflows
- Canva — for content creation workflows
- Notion — for knowledge bases, docs, and task sync
Caveat about incomplete MCPs
The host notes that some MCPs expose only part of a product’s functionality. For example, he says Beehive’s MCP may currently provide data access, but not full posting functionality yet.
How to Add an MCP
The process described is simple:
- Find the MCP URL for the service you want.
- In Claude or another assistant, choose “Add your own connection.”
- Paste the MCP link.
- Log in through the authentication pop-up.
- Start giving the assistant tasks inside that connected tool.
The host specifically says this works well with AI Box and other services that provide their own MCP links.
Best Use Cases for MCPs
MCPs are especially valuable when you need:
- Repeated generation tasks
- Multi-step pipelines
- Workflows that must not lose output
- Access to tools the AI doesn’t natively support
- More efficient automation with fewer token costs
Examples mentioned:
- Generating ad images
- Building newsletters
- Creating podcast assets
- Producing long-form audio content
- Creating Bible study explanations at scale
- Syncing work across platforms like GitHub, Supabase, and Vercel
Key Takeaways
- MCPs are a standardized way for AI assistants to use external tools.
- They are often easier and more reliable than raw API integrations.
- They can unlock new capabilities like image, audio, and video generation inside Claude, ChatGPT, or Gemini.
- They are especially useful for large-scale, repetitive, or failure-sensitive workflows.
- A strong starter stack for builders is GitHub + Vercel/Netlify + Supabase.
- If your software has an MCP, it is usually worth connecting it.
Mentioned Resource
- AI Box MCP: The host says it costs $8/month and can be found at aibox.ai/mcp
- He says it lets users connect Claude, ChatGPT, or Gemini to AI Box’s tools and access 80+ AI models
Bottom Line
This episode is essentially a case for making AI assistants more useful through MCP integrations. The host argues that MCPs reduce friction, improve reliability, and unlock powerful new workflows—especially if you use Claude, ChatGPT, or Gemini for real work.