2nth builds the skills layer for AI agents — structured, agent-loadable knowledge that lets businesses deploy AI that compounds instead of running pilots that stall.
One knowledge tree. Read by humans, loaded by agents, operated as a system. Built in Johannesburg for businesses that need AI grounded in their domain — and their regulations.
Most organisations are running disconnected pilots — chatbots, copilots, one-off workflows. Activity without progress: experiments that don't connect, don't improve, and don't reuse what came before.
The fix isn't another tool. It's a harness: structured skills your agents load, run, and improve — instead of headcount-shaped AI bolted onto org charts.
harness, not headcount
We map your work into outcomes, roles, skills, and tasks — then deploy agents into your actual environment. CRM, ERP, data, comms. Not a sandbox.
Every skill node is structured context your agent can load today. Agent Skills format — native in Claude Code, Claude.ai, and the API.
Your domain expertise, operationalized as a private branch — co-branded nodes, agent-ready context packages, institutional knowledge that compounds instead of walking out the door.
Twelve domains. One tree. Every node serves two readers — the human learning and the agent operating. This isn't a content library; it's the context layer your agents run on.
Skill nodes are plain SKILL.md with progressive disclosure — load a branch into your agent's context and it starts domain-aware. No SDK, no lock-in.
Point Claude Code, Claude.ai, or the API at a node and it reads like documentation and loads like context.
Browse all 12 domains →# Point your agent at a live skill node curl https://know.2nth.ai/explainers/agents/skills.html # …or clone the public skills tree into Claude Code git clone https://github.com/2nth-ai/skills # a node is just SKILL.md + progressive disclosure: --- name: agent-skills description: Author and load Agent Skills that Claude reads as docs and loads as context. domain: agents --- # Agent Skills A skill is structured, agent-loadable knowledge…
We turn your work into outcomes, roles, skills, and tasks.
Each skill becomes an agent-loadable node on the tree.
Agents run those skills in your real stack — CRM, ERP, data, comms.
Every run sharpens the skills the next run loads.
Every run generates data. Data sharpens skills. Skills compound across the tree. That's the 2ⁿ effect — scale by capability, not headcount.
We published a cross-page audit correcting three of our own overstated SA-residency claims, with primary sources. That's the standard every node on the tree is held to.
Read the SA LLM Residency validation memo →A professional-services team turned a manual proposal process into a mapped skill an agent runs in their own stack.
Anonymized at the client's request · specifics on a call
POPIA-aware data handling. FSCA and SARB context in finance nodes. NHBRC and JBCC in construction. ICASA in IoT. Load-shedding-resilient edge patterns. AI for African businesses can't be a US template with the logo swapped — the tree encodes the local layer.
Customer, developer, or partner — tell us which door you're coming through and we'll route you to the right next step.
