This weblog put up focuses on new options and enhancements. For a complete record, together with bug fixes, please see the launch notes.
We’re rolling out two key options that change the way you construct AI utilizing Clarifai: help for AI brokers and the Mannequin Context Protocol (MCP).
AI Brokers: Constructing Smarter, Autonomous AI
AI brokers are an enormous step past single-task AI fashions. As an alternative of simply doing one factor, an agent can purpose, plan, and take a number of actions to realize a bigger purpose. Consider them as AI packages that may break down complicated issues and use totally different instruments or fashions to get the job achieved.
With this launch, we’re making it simpler to construct these brokers on Clarifai. This implies you may:
- Create goal-oriented AI: Design techniques that work in the direction of particular targets, not simply offering remoted solutions.
- Chain collectively AI capabilities: Mix a number of fashions and instruments on our platform (or exterior ones) in sequence to unravel extra complicated issues.
- Automate multi-step processes: Cut back guide effort by having AI deal with complete workflows.
This opens up potentialities for extra superior AI functions that may make selections and adapt to conditions.
To point out you what and how one can construct AI Brokers, we have created an AI Weblog Writing Agent utilizing Clarifai and CrewAI!
On this video, we construct an AI-powered weblog writing agent that generates full weblog posts from scratch. We use:
- CrewAI to handle agent orchestration
- Gemini 2.5 Professional mannequin powered by Clarifai
- Streamlit to create a easy and interactive UI
MCP: Giving AI Brokers Actual-World Context
For AI brokers to be really helpful, they want entry to real-time info from outdoors their inner information. The Mannequin Context Protocol (MCP) solves this by offering a standardized manner for AI fashions and brokers to work together with exterior information sources and APIs.
We have built-in MCP, permitting you to:
- Join brokers to your information: Bridge your AI brokers along with your firm’s databases, information lakes, and different inner techniques.
- Entry dwell information: Give your brokers present info from exterior APIs, like monetary information, information, or sensor readings.
- Construct customized information bridges: Create your personal MCP servers to tailor how your AI brokers entry and use exterior context.
Combining AI brokers with MCP means your AI can’t solely suppose and plan but in addition actively fetch and use real-world info, making your AI functions extra highly effective and related. Be taught extra right here.
Clarifai now provides an OpenAI-compatible API endpoint, permitting you to make use of your current OpenAI code and workflows to run inferences with Clarifai fashions, together with people who combine or wrap OpenAI fashions.
The compatibility layer routinely interprets OpenAI-style requests into Clarifai API calls, so you may entry Clarifai’s broad mannequin library as customized instruments inside your OpenAI-based initiatives.
This removes the necessity to rewrite your code for Clarifai’s native API, making integration quick and easy for groups already acquainted with OpenAI.
Under is an instance that makes use of the OpenAI Python consumer library to work together with a Clarifai mannequin by way of Clarifai’s OpenAI-compatible API endpoint. Learn extra right here
We’ve made quite a few enhancements to the Python SDK to boost stability, usability, and integration capabilities:
We’re excited concerning the new Agentic and MCP help in Clarifai and are trying ahead to seeing the sorts of functions the neighborhood builds round it. Try our video tutorial on constructing an AI Weblog Writing Agent to see AI Brokers in motion. You may also discover extra examples right here.
Discover the documentation and begin constructing as we speak. We’ll even be including extra agent examples and templates quickly, so keep tuned.
You probably have any questions, ship us a message on our Group Discord channel. Thanks for studying!