The full agentic capability library

Goal-driven, supervised AI agents that run whole workflows end-to-end — with the guardrails, oversight and audit trail that make autonomy safe to trust.

techniques

disciplines

capabilities

the whole spectrum

assisted → autonomous · not just the fashionable end

20 assisted34 supervised31 autonomous

/ browse

The 5 capabilities, in depth

Pick a capability — its plain-English job, the questions it answers, and the full toolbox scroll alongside.

the colour is the tierassistedsupervisedautonomous

Reason & plan

break the goal into the right steps

17techniques · 2 disciplines

4 assisted7 supervised6 autonomous

Give the agent the ability to think a problem through, decide what to do, and reach for the right tool — instead of answering in one shot and hoping it's right.

answers questions like

→“Can it break a fuzzy goal into the right sequence of steps on its own?”
→“When it needs data, code or a system action, does it pick and call the right tool?”
→“When it's wrong, can it notice and correct itself before you see the answer?”

the toolbox — 17 techniques, assisted to autonomous

Reasoning patterns

10 techniques

Single-promptChain-of-thoughtReAct (reason + act)Plan-and-ExecuteReflection / self-critiqueSelf-consistencyTree-of-ThoughtsPlanner-workerLanguage Agent Tree Search (LATS)Reasoning models (extended thinking)

Tool use

7 techniques

Manual copy-pasteTemplated single callTool / function-callingStructured (JSON) outputsRetrieval as a tool (GenAI grounding)Code-act (code as actions)Computer-use / browser-use

/ the point

Autonomy needs a control plane

Gartner projects 40%+ of agentic-AI projects will be cancelled by 2027 for weak controls, unclear value and runaway cost — so we ship the AgentOps guardrails, approvals and audit trails that make autonomy safe to trust.

Back to the service

models & RAG grounding with Generative AI · serving & monitoring with AIOps / MLOps · governance & operating model with AI Strategy & Advisory