Introduction
Time series power decisions across energy, finance, retail, and beyond. In recent years, foundation models have transformed how we approach time series problems, making forecasting and analysis faster, more powerful, and more accessible than ever.
Despite this progress, teams still face a fragmented ecosystem: different models live in different tools, workflows are brittle, and applying intelligence beyond single predictions remains hard. What's missing is a unified platform that brings together leading foundation models, understands natural language, and can reason about time series data end to end.
Today, we're excited to introduce a major step toward that vision at Nixtla.
What's New in Nixtla Enterprise
This release introduces three major capabilities that together expand Nixtla from a single-model offering into a full time series intelligence platform:
- Access to leading foundation models beyond TimeGPT
- A standardized interoperability layer via Model Context Protocol (MCP)
- Native agentic capabilities for end-to-end analytical workflows
Together, these unlock a new way to build, compare, and operationalize time series intelligence.
Access Multiple Foundation Models from a Single Interface
Nixtla Enterprise now lets you work with multiple leading time series foundation models directly from the NixtlaClient. In addition to TimeGPT, you can now access models such as Chronos and TimesFM through a unified interface.
This makes it easy to:
- Experiment across different model families
- Benchmark performance on your own data
- Select the right model for each use case
All without switching tools, managing separate APIs, or rewriting pipelines.
Introducing MCP & Agentic Capabilities for Time Series Workflows
To support this multi-model ecosystem, Nixtla now integrates with Model Context Protocol (MCP). MCP provides a standardized way for models and agents to access time series data, tools, and execution environments.
Rather than hard-coding workflows around individual models, MCP enables:
- Consistent access to datasets and metadata
- Reusable analytical tools across models
- A shared context for reasoning and execution
Nixtla’s MCP tools work seamlessly with Claude Desktop, Claude Code, and Cursor, and our dedicated installers make them safer and easier to use in production and local environments.
MCP acts as the connective tissue that allows models and agents to work together seamlessly inside Nixtla. Built on top of multi-model access and MCP, Nixtla now supports agentic workflows for time series analysis.
Agents can:
- Interpret natural language instructions
- Reason about time series structure and objectives
- Execute code in a fully hosted environment and generate business insights
- Evaluate results and iteratively refine solutions
This moves beyond static forecasting APIs toward autonomous analytical systems that can operate end to end — from exploration to production-ready insights.
Real-World Results
We've been testing these capabilities in real settings, including the VN1 forecasting competition. Using a combination of multi-model access and agentic reasoning, Nixtla's system autonomously generated and refined forecasting solutions with minimal human guidance.
The result: first place, outperforming the original competition winner and previously published solutions. This demonstrates how combining foundation models with intelligent control and execution can unlock substantially better outcomes in time series tasks.
Looking Ahead
This release marks an important milestone in Nixtla's mission to build an inclusive and autonomous time series ecosystem. By bringing together leading foundation models, a standardized intelligence layer, and agentic capabilities, we're enabling a future where time series analysis is not only automated, but adaptive.
This is just the beginning.