New Product Release: TeamSpark AI Workbench - April 28, 2025

Build Your Own Intelligent Agents with TeamSpark AI Workbench

We've all experienced the power of large language models through chat interfaces. They can answer questions, draft text, and even write code. But what if you want to go beyond simple back-and-forth? What if you need an AI that can remember specific information, follow complex instructions, and interact with the world using tools? Building truly intelligent, task-oriented AI agents often feels like a daunting technical challenge, requiring deep knowledge of APIs, prompt engineering, and complex orchestration. That's why we built TeamSpark AI Workbench.

TeamSpark AI Workbench is a desktop and command-line client application for Mac, Windows, and Linux that allows anyone to easily build and control intelligent agents tailored to their specific needs. It provides the core building blocks you need to give your AI agent memory and specialized knowledge, teach it how to think, and equip it with the ability to act.

Rather watch a product demo video than continue reading?

For those of you still with us…

Let's look at the key components that make this possible:

References: Your Agent's Memory

Just like you rely on notes, documents, or your own memory to tackle a problem, your AI agent needs information to draw upon. In TeamSpark AI Workbench, References are how your agent collects and remembers information. A Reference can hold any kind of information: text you write, content fetched from a document or webpage, summaries of previous interactions, or results from tool calls. It's essentially any piece of knowledge you need to provide to your agent. Each Reference has a name, a description, a priority level, and settings for when it should be included in the agent's context (always, on demand, or manually). This gives you precise control over the information your agent has access to, and ensures that your agent is prioritizing the most important and relevant information.

TeamSpark AI Workbench provides tools to easily create and manage these references. Plus, with the optional internal References tool, the AI itself can be instructed to create, modify, or include references based on your commands, so you can tell your agent things like:

• "Summarize the findings from this chat session in a new reference"
• "Get the web page at http://somewebsite.com/document and convert its contents into a new reference"

Rules: Guiding Your Agent's Behavior

If References are your agent's memory, Rules are its guiding principles and problem-solving techniques. Rules govern the way your agent instructs the AI to process information and commands. Rules are essentially specialized "system prompts" injected into each interaction to help the AI respond in the correct way. This is where you apply prompt engineering techniques (though plain language works pretty well) to teach your agent how to approach different types of tasks or solve specific kinds of problems. Rules are how an agent "learns" problem-solving skills. As with references, you have complete control over when rules are active and in use.

With the optional internal Rules tool, you can even instruct the AI to create or modify rules based on your interactions, enabling scenarios like:

• "Make a new rule so that you always output properties with attributes in markdown table form"
• "Make a new rule for yourself that will ensure that when we encounter a challenge like this in the future you will process it using the technique that was successful above"

Tools: Extending Your Agent's Capabilities

An intelligent agent isn't just a conversational partner; it needs to interact with the real world. Tools are specialized functions or capabilities that your AI agent can use to perform specific tasks, extending its abilities beyond just conversation. TeamSpark AI Workbench agents can use tools to:

• Search and fetch content from the internet
• Read from or write to your file system (within allowed directories)
• Execute database queries
• Run system commands
• Get real-time information like weather forecasts
• And much more, depending on the tools you add!

Tools are automatically available to the AI when processing messages, allowing it to gather information, perform calculations, and automate tasks to accomplish your goals more effectively.

TeamSpark AI Workbench uses the Model Context Protocol (MCP) to manage and interact with tools. MCP is an open protocol that standardizes how applications provide context and capabilities to large language models. It's like a universal adapter for connecting AI models to external functions and data sources. You can learn more about MCP and explore the growing collection of MCP servers at https://modelcontextprotocol.io

Powered by Your Choice of AI

TeamSpark AI Workbench is designed to be flexible. It allows you to use AI models from a variety of providers, including Anthropic (Claude), OpenAI (ChatGPT), Google (Gemini), AWS Bedrock, and Ollama. You choose the model that best suits your task, budget, and security and compliance needs.

Get Started Building Your Own Agents

TeamSpark AI Workbench puts the power of building sophisticated AI agents into your hands. By combining References, Rules, and Tools, you can create custom agents that remember context, follow specific instructions, and interact with the world to help you solve complex problems. Ready to stop just chatting with AI and start building with it? Visit our website at https://www.teamspark.ai to learn more and download TeamSpark AI Workbench for your operating system.