Clarifai 11.5: Introducing Support for AI Agents and Model Context Protocol (MCP) | Clarifai Blog

.png)

.png?width=1000&height=556&name=11.5_blog_hero%20(1).png)

This blog post focuses on new features and improvements. For a comprehensive list, including bug fixes, please see the release notes.

Introduced support for agentic frameworks

We're rolling out two key features that change how you build AI using Clarifai: support for AI agents and the Model Context Protocol (MCP).

AI Agents: Building Smarter, Autonomous AI

AI agents are a big step beyond single-task AI models. Instead of just doing one thing, an agent can reason, plan, and take multiple actions to achieve a larger goal. Think of them as AI programs that can break down complex problems and use different tools or models to get the job done.

With this release, we're making it easier to build these agents on Clarifai. This means you can:

• Create goal-oriented AI: Design systems that work towards specific objectives, not just providing isolated answers.

• Chain together AI capabilities: Combine multiple models and tools on our platform (or external ones) in sequence to solve more complex problems.

• Automate multi-step processes: Reduce manual effort by having AI handle entire workflows.

This opens up possibilities for more advanced AI applications that can make decisions and adapt to situations.

To show you what and how you can build AI Agents, we've created an AI Blog Writing Agent using Clarifai and CrewAI!

In this video, we build an AI-powered blog writing agent that generates complete blog posts from scratch. We use:

• CrewAI to manage agent orchestration

• Gemini 2.5 Pro model powered by Clarifai

• Streamlit to create a simple and interactive UI

MCP: Giving AI Agents Real-World Context

For AI agents to be truly useful, they need access to real-time information from outside their internal data. The Model Context Protocol (MCP) solves this by providing a standardized way for AI models and agents to interact with external data sources and APIs.

We've integrated MCP, allowing you to:

• Connect agents to your data: Bridge your AI agents with your company's databases, data lakes, and other internal systems.

• Access live data: Give your agents current information from external APIs, like financial data, news, or sensor readings.

• Build custom data bridges: Create your own MCP servers to tailor how your AI agents access and use external context.

Combining AI agents with MCP means your AI can not only think and plan but also actively fetch and use real-world information, making your AI applications more powerful and relevant. Learn more here.

Introduced the OpenAI-compatible API endpoint

Clarifai now offers an OpenAI-compatible API endpoint, allowing you to use your existing OpenAI code and workflows to run inferences with Clarifai models, including those that integrate or wrap OpenAI models.

The compatibility layer automatically translates OpenAI-style requests into Clarifai API calls, so you can access Clarifai’s broad model library as custom tools within your OpenAI-based projects.

This removes the need to rewrite your code for Clarifai’s native API, making integration fast and simple for teams already familiar with OpenAI.

Below is an example that uses the OpenAI Python client library to interact with a Clarifai model via Clarifai's OpenAI-compatible API endpoint. Read more here

Published Models

• The latest DeepSeek-R1-0528-Qwen3-8B is now available on the platform. This new model builds on DeepSeek-R1 v2 with significant enhancements in reasoning and logic, driven by more efficient computation and optimization. It approaches the performance levels of leading models such as o3 and Gemini 2.5 Pro.

• Published Qwen2_5-Coder-7B-Instruct, a code-specific LLM series (0.5B–32B) with improved code generation, reasoning, and fixing. Trained on 5.5T tokens, the 32B model rivals GPT-4o in coding capabilities.

• Published Claude Opus 4, a state-of-the-art large language model from Anthropic. It supports text and multimodal inputs and can generate high-quality, context-aware text completions, summaries, and more.

Token-Based Billing

• The token-based billing will align pricing with industry standards and better reflect the costs associated with these models.

• The token-based pricing will be gradually applied only to inference on the Community using models deployed with Clarifai's default Shared compute. Dedicated compute instances will continue to use per-compute-time billing, regardless of the type of model deployed.

• Also, the per-request billing will continue for legacy vision models on the Community.

Python SDK

We have made numerous improvements to the Python SDK to enhance stability, usability, and integration capabilities:

• Restored support for pretrained model configuration files.

• Added the clarifai model init CLI command to generate default files for model uploads.

• Resolved issues affecting model upload reliability and workflows.

• Improved handling of Clarifai configuration during URL construction.

• Updated code snippets for MCP and OpenAI integrations.

• Fixed a bug in MCPModelClass that impacted notifications.

• Enhanced OpenAIModelClass to streamline request processing, improve modularity, and simplify parameter extraction and validation.

• Fixed a bug in OpenAIModelClass to ensure full JSON responses are returned.

• Cleaned up fastmcp implementation for better maintainability.

• Added OpenAIModelClass to support OpenAI-compatible API endpoints.

• Fixed utility functions for OpenAI messages and code snippets.

There are many more updates. Learn more here.

Additional Changes

• The Node.js SDK now includes a more efficient model inference method and supports OpenAI-compatible API endpoints. We fixed type export issues, ensured compatibility with both ESM and CommonJS, and addressed security vulnerabilities by updating packages. Learn more here.

• Clarifai Organizations improved usability by hiding the create button for users without permissions and showing a persistent invite popup until users respond, boosting invite visibility.

• In the Control Center, chart colors are now consistent across reports, a clear “no data available” message replaces zeros when data is missing, and the Teams and Logs tab is hidden from users without audit logging access. Learn more here.

• Platform updates added a "Remember Me" option to the homepage login, improved cover image generation, introduced pagination for personal access tokens, refined homepage usability, hid Recent Activity for unauthorized users, and enhanced the billing interface for a smoother credit card and plan management experience.

Ready to start building?

We’re excited about the new Agentic and MCP support in Clarifai and are looking forward to seeing the kinds of applications the community builds around it. Check out our video tutorial on building an AI Blog Writing Agent to see AI Agents in action. You can also explore more examples here.

Explore the documentation and start building today. We’ll also be adding more agent examples and templates soon, so stay tuned.

If you have any questions, send us a message on our Community Discord channel. Thanks for reading!