Cohere’s North 2 AI Platform Adds Third-Party Model Integration
In a strategic move designed to give businesses greater control over their artificial-intelligence deployments, Cohere has introduced North 2, an updated version of its AI-agent harness platform that now supports integration with models created by competing developers. The Toronto-based firm says the new system enables organizations to set clear usage limits, plug in third-party AI models, and tap pre-built tools that are meant to accelerate adoption across teams that lack deep technical expertise.
Expanded Capabilities for Enterprise AI Management
North 2 lets staff at deploying organizations configure AI agents that can handle longer-term and more detailed tasks. The updated system will hold on to more details of those assignments in an attempt to improve automations over time.
The platform also introduces budget controls, allowing companies to define spending caps for individual employees or entire teams. In addition, it offers tracking features that help measure the return on investment generated by AI-powered workflows.
According to Cohere’s vice-president of product for AI, Paul Teyssier, the update marks a shift from experimental AI use toward production-grade deployments at scale. North was initially released in January 2025 with the Royal Bank of Canada as a flagship client, later expanding to include Bell, Saudi telecom operator STC Group, Canada’s Innovation Department, and the Aston Martin F1 team.
Interoperability as a Strategic Differentiator
Unlike other AI-agent harnesses such as Anthropic’s Claude Code and OpenAI’s Codex, North 2 permits customers to integrate models from a range of providers, including Anthropic, OpenAI, Google, and Chinese firms like Alibaba, DeepSeek, and Z.ai. This modular approach gives clients the freedom to select the model that best fits a particular task, whether that means leveraging cutting-edge systems from major U.S. players or opting for cost-effective open-source options from Asia.
“A whole bunch of players are trying to take the customers deeper and deeper into their ecosystem, with no way out,” Teyssier said. “We are intentionally taking the other approach.” This interoperability could reduce reliance on the company’s own models, which have been open-sourced to allow free on-premise deployment. Nevertheless, the firm continues to generate revenue through hosted model access and premium offerings.
Competing in a Crowded AI Tools Market
While the technology may not match the coding and cybersecurity strengths of rivals like Anthropic and OpenAI, the company is focusing on specialized applications that matter to regulated industries such as finance, telecom, defense, and government. It has also built models optimized for business tasks like database retrieval and document extraction from scanned materials.
The firm expects most clients to keep using its models for routine AI applications, reserving alternative systems for niche use cases or experimentation. “They’re going to prefer our models based on hard facts, not based on the fact that it’s locked together,” he explained.
In August, fintech company Stripe acquired OpenRouter for a reported US$7.5 billion, highlighting the growing demand for such solutions.
Cost management has become the top concern for AI leaders, according to Teyssier. North 2 addresses this by enabling automatic downgrading to less expensive tools for simpler tasks and by providing granular budget tracking. Cohere hasn’t yet built its own router into the platform, but is seeing demand for it, Teyssier said.
