French AI startup Mistral AI has introduced its Agents API, a new framework built to help enterprises create AI agents that perform tasks, maintain long-term context, and coordinate multi-step workflows.
According to Mistral, the Agents API goes beyond the capabilities of traditional language models, which primarily focus on generating text.
“By providing a reliable framework for AI agents to handle complex tasks, maintain context, and coordinate multiple actions, the Agents API enables enterprises to use AI in more practical and impactful ways,” the company said in a blog post.
The API integrates with built-in connectors for code execution, web search, image generation, and document access. It also supports tools that follow the Model Context Protocol (MCP), a standard for connecting AI agents with external systems such as APIs and databases.
The company said the API supports persistent memory across conversations, allowing for stateful interactions. This means developers can view and continue past conversations, or start new ones from any point. The API also supports streaming outputs.
The orchestration feature lets developers build workflows involving multiple agents. For example, a financial agent can assign tasks to a web search agent or a calculation agent.
“A single request can trigger tasks across multiple agents, each handling specific parts of the request,” the company said. “This collaborative approach allows for efficient and effective problem-solving, unlocking powerful possibilities for real-world applications.”
Mistral showcased use cases across several domains, including a coding assistant that interacts with GitHub, a financial analyst agent that compiles metrics, a travel assistant, and a nutrition planner. Each agent operates through defined workflows, often using multiple MCP tools and connectors.
In benchmark tests, Mistral’s models with web search support showed marked improvement. The company cited its SimpleQA (OpenAI’s benchmark) results, where Mistral Large with web search scored 75% compared to 23% without it.
The company has made documentation and a set of cookbooks available to developers. Mistral’s Agents API complements its existing Chat Completion API and is positioned as a core building block for enterprise-grade agentic platforms.
Capgemini recently announced a three-way partnership with Mistral AI and SAP to deliver scalable generative AI solutions tailored for highly regulated industries such as finance, aerospace, and the public sector.
The initiative aims to unlock AI-driven business value while maintaining strict data security and compliance standards. By combining Mistral AI’s multilingual models with SAP’s secure Business Technology Platform, Capgemini will offer over 50 pre-built AI use cases across sectors.
Mistral has recently released several new models, including Mistral Medium 3, Devstral for coding, Mistral Document AI (OCR), and Mistral Saba for South Asian languages.