MCP is a protocol built to connect language models with external systems and data sources, standardizing the way these models access information and perform tasks. It lets a language model expand its abilities by using context and data from multiple sources to come up with more accurate and relevant answers.
Imagine you have a virtual assistant on your phone and you want it to help organize your schedule, check the weather, and even make restaurant reservations. Without MCP, the assistant would need to talk separately to each service (calendar, weather forecast, reservation system), which can lead to inconsistency and slowdowns. With MCP, the assistant can access all these systems in a standard and integrated way, process everything together and give you a complete and clear answer, like "The meeting is set for 2pm, today's weather is sunny – plus, I found a great restaurant with a dinner spot available." That makes things feel much more natural and smooth, even for someone who isn't super technical.
The installation process for the MCP to integrate with TESS is straightforward and simple. Using Claude Code, you just need to provide the local MCP name and Tess's API key.
Installation is done through the npx command, installing the package with the command below
npx -y mcp-tessIt's important to note that the API shown in the video above was discontinued after the recording, and is only used for educational purposes.
Once it's installed, Claude Code is ready to interact with Tess AI's endpoints!
The integration between TESS and MCP makes all the endpoints in the TESS API easier to access. The official TESS AI documentation gives you detailed info about each available endpoint, including how to generate access tokens, run operations, contract versioning, and other key features.
What makes this integration special is the abstraction layer that Claude Code adds, letting you interact with these endpoints in a more natural and intuitive way, without having to deal directly with complex HTTP calls
TESS Memory System
TESS AI has a robust memory system that lets you store and organize important info. In the example shown in the video, at first there’s only a default collection called "MyMemories" with no memory stored. The memory system works like a persistent knowledge bank that any TESS agent can access and use during its runs.
Creating a New Collection
The first practical step shown is creating a new collection called "Boas Práticas de Programação". Using Claude Code, the command is naturally interpreted and the system asks if you want to use TESS’s MCP to run the operation.
After you confirm, the collection gets created successfully and you can check it either through listing commands or right in the TESS AI web interface.
Adding Memories to the Collection
Once the collection is created, the next step is to add relevant memories. In this example, two important memories about programming best practices are added:
"Always use SOLID as a standard for development"
"Always create unit tests using BDD"
Each memory is added individually using natural commands in Claude Code, and the system confirms the success of each operation. The memories are immediately available in the collection and can be checked either through list commands or in the TESS web interface.
To show the real power of the integration, a more complex challenge is suggested: create a flight reservation management service in Node.js, using the best practices stored in the collection built earlier.
The cool thing is that the command doesn't have to be super specific – just mentioning "Collection de Boas Práticas" (without the full name) and the agent "TES AI V5" is enough for the system to get the context.
Claude Code reads the command and runs a series of steps:
Agent Search: First, the system searches for the specified agent "TES AI V5", which is the main TESS AI chat agent.
Prompt Preparation: The system builds a full prompt including:
The request to create the reservation management service
The technical specs (Node.js)
The reference to the memories from the collection de boas práticas
Execution and Generation: The agent handles all this info and creates a full setup for the requested service.
The agent shows it understood and applied the instructions correctly by:
Creating a code structure following the SOLID principles stored in the memories
Including unit tests using BDD, as specified in the second memory
Developing a complete and functional flight booking management service
This proves the system not only accessed the memories, but also interpreted and applied them contextually and intelligently when generating code.
This integration marks a big step forward in how we interact with AI systems. By combining the natural language processing power of Claude Code with the strong features of TESS AI, we create an environment where:
Developers can interact with complex APIs using natural language
Knowledge and best practices can be stored and reused automatically
AI agents can access and apply contextual information in a smart way
The technical barrier to using advanced AI features is significantly reduced
This demo shows not just the technical features, but also the game-changing potential of this integration to speed up the development of AI-based solutions, making them more accessible and efficient for developers of all experience levels!