Cohere North 2 launched on October 5, 2026, the biggest upgrade yet to the company's enterprise agentic AI platform, adding cross-session agent memory, a redesigned orchestration harness, and granular administrative cost controls, according to coverage of the announcement by Unite.AI. The release is aimed squarely at regulated industries that want AI agents running on infrastructure they control rather than in a vendor's cloud.

Cohere describes North 2 as unifying enterprise-grade security, intelligence, cost governance, and full-stack control in a single agentic platform that runs wherever an organization's data lives, according to the company's announcement. The company said a year of production deployments shaped the new version, with rollouts across finance, healthcare, telecommunications, manufacturing, energy, and the public sector feeding back into the design.

Memory and skills for reusable enterprise agents

At the core of Cohere North 2 is what the company calls an advanced agent orchestration system for complex, multi-step processes. According to the announcement, users can build agents and automations from simple prompts while keeping humans in the loop, and reusable agents can be shared across an entire organization instead of being rebuilt by every team.

The release introduces several named feature groups. Skills are reusable capabilities agents can call on so employees do not rebuild the same logic, while Libraries hold shared knowledge and assets that the whole organization can draw from. Memory lets agents retain context across sessions instead of starting fresh each time. Applications lets users prototype decks, dashboards, documents, and lightweight apps from natural language, and Automations provides pre-built templates, a drag-and-drop visual workflow builder, and real-time performance monitoring.

New enterprise connectors link North to workplace tools including Slack, SharePoint, OneDrive, Microsoft Outlook, Microsoft Exchange, Jira, Linear, Notion, and GitHub. Financial data connectors to PitchBook, Crunchbase, Daloopa, FiscalAI, S&P Global, and FactSet are listed as planned, according to the announcement.

One adopter cited in the announcement is LG CNS, which uses North to bring AI to customers at scale across South Korea. Yohan Jin, the company's Head of AI Center, said in the announcement that North gives enterprises a secure, practical foundation for turning agentic AI into real business value, and that the firm's experience deploying it within LG CNS and with customers across Korea has strengthened confidence in scaling intelligent automation across the enterprise, as reported by Unite.AI.

Security, sovereignty, and cost governance

North 2 deploys privately on-premises, in the cloud, or fully air-gapped, with environment options including self-hosted, virtual private cloud, hybrid, and on-premises configurations. Cohere lists SOC 2 Type 2, ISO 27001, and ISO 42001 certifications for the platform.

According to the company, North Admin lets teams enforce guardrails at the individual agent level, filtering content, improving safety, and validating model responses. Autonomy policies restrict agents to actions they are authorized to take and require human oversight for critical decisions. Cohere said it conducts continual security testing, including red-teaming exercises and third-party vulnerability scanning, and provides observability tools that log all changes. Bell Cyber is partnering with North to bring government-grade security to agentic AI, the company said, and Bell Cyber's chief technology and AI officer said the platform lets his organization apply agentic AI inside security operations while maintaining the control, privacy, and data sovereignty standards its customers require.

North Admin also supplies the release's governance layer. Administrators can set roles and permissions, control which models are used where and by whom, and monitor activity in real time with detailed analytics on token spend. Alert thresholds can notify teams before limits are reached, and a flow-control feature defines consumption tiers based on requests and token rates for users and groups. The platform is model agnostic, so organizations can run Cohere's generative models or bring their own existing models.

Cohere also reported findings from its collaboration with NVIDIA: North generates more insights for fewer tokens when Cohere models are trained and deployed on NVIDIA Blackwell and Hopper hardware, and running its models on NVIDIA hardware increases tokens per second per node, effectively decreasing spend, according to the announcement.

The enterprise agent platform race

Cohere North 2 arrives amid an intense race to become the default platform for enterprise AI agents. Just days earlier, IBM announced that its Bob agentic software development platform can now run entirely self-hosted for banks and governments unwilling to send proprietary code to a third-party AI cloud — a story covered when the news broke. Meanwhile OpenAI's DevDay headlined Dots, always-on agents powered by GPT-6 Astra, the model behind the ultrafast Blackwell deployment reported last week.

The through-line in these launches is control: enterprises want the productivity gains of agentic AI without surrendering data sovereignty, security posture, or cost visibility. Cohere North 2's bet is that memory, model-agnostic deployment, and granular cost governance are what unlock production use at scale — not bigger models alone. For more on the agent ecosystem, see the site's AI News coverage.

Cohere first made North widely available in August 2025 as an agentic AI platform that securely accesses workplace data with private deployment, and the company points to customers including RBC, Dell, LG CNS, and Bell from that era. With North 2, it is asking enterprises to judge agentic platforms the way they judge any infrastructure: by how securely, cheaply, and controllably they run, not by how impressive the demo is. According to Cohere's announcement, organizations interested in North 2 can book a demo through the company's sales team.