Introducing North 2. Enterprises have been compromising on agentic AI for years: platforms that aren't scalable, models that aren't sovereign, control without enterprise-grade security. That ends today, with 15+ new features in our biggest upgrade yet:

Sovereign: Control where and how your models of choice run, with the ability to prototype documents, on-prem and air-gapped deployments, and shared knowledge for the whole org.

Synchronized: Connect North to everyday apps such as Slack, Sharepoint, OneDrive, Outlook, Exchange, Jira, Linear, Notion, and GitHub to make your work go faster than ever.

Secure: We go above and beyond global security standards, with SOC 2 Type 2, ISO 27001, and ISO 42001 certifications, plus easy-to-implement guardrails.

Smarter: With North 2, you can create reusable agents and automations and multi-agent orchestrations so each agent can swiftly complete your asks. Plus, agents now keep context across sessions, instead of starting cold every time.

Successful enterprises focus on actually getting work done. We built North 2 for them. No flashy headlines, no overpromising. Just things that work. Book a demo today:

I thought this was a new north mini code model and I got excited for a second.

idk man, i don't know anything about your product but having more than one north doesn't seem like an improvement. work on the naming?

Surprised this company is still around Literally don't know anybody using cohere in enterprise

Sovereignty is usually the first thing cut from agent pilots. Glad it leads here.

21% of enterprises still only monitor agent spend reactively, with no programmatic stop. Hard token caps per user and per agent in North 2 close that gap. Same discipline belongs in trading agents: a budget and a kill switch before the loop can spend or trade without a gate.

Sovereignty is the selling point now, so the real test is whether Cohere's own models hold up without a fallback to the frontier labs.

Context across sessions is the real upgrade here. Cold starts are where agents lose the most time. Is that memory shared across the multi-agent setup or kept per agent?

想快速验证 North 2 值不值得试点，可以先做一个20分钟测试：①拿自己最痛的一个内部工作流（比如跨系统查数），看它能不能一次跑通而不是拆成多步人工；②把同一个任务丢给现有方案，对比端到端耗时和出错率；③故意给一个缺权限/缺数据的场景，看它的拒绝和升级路径清不清晰。

a per-user, per-agent spending cap sounds boring. it's the part that matters most. an AI that can spend company money needs a budget, like any new hire.

No flashy headlines, but 'that ends today' is doing some flashing

Per-agent token caps are the right primitive. The next useful layer is tying each cap to a named task, so an agent that hits its limit stops and hands back a partial result instead of quietly retrying.

This slop video is not acceptable for a tasteful company like yours.

open weights or it didn't happen

Cohere North 2’s focus on sovereign, secure, and scalable enterprise AI is a strong step forward. I’d love to collaborate and help amplify this launch through creator-led content and marketing. Open to discussing a partnership?

15 new features in one drop is a lot. Which one were enterprise teams asking for loudest?