An enterprise AI agent platform for AI automation and governance: visual workflow builder, private knowledge bases with retrieval-augmented generation, data preparation and data manipulation steps, human approval, AI Guardians, credential vault, cost control and multi-cloud model access.
ATX Agent Hub is where an organisation builds, tests, approves and runs AI agents. It combines a no-code workflow builder, private knowledge bases, multi-model access and the governance controls that regulated teams need before an AI agent is allowed to touch real work.
An AI agent here is a workflow: a sequence of steps that reads from your systems, reasons with a language model over your own documents, and acts — creating a Jira issue, drafting an email, updating a sheet, calling an API — with a person approving the steps that matter.
Knowledge base collections index PDFs, Word documents and pages from Confluence or SharePoint into vector collections. Agents and the Chat with Data assistant answer from that content through retrieval-augmented generation, with a configurable citation style, a confidence score and document lineage down to the chunk that was used.
Every workflow has immutable versions. Promotion to production needs an independent approver, and requesters cannot approve their own changes. Project roles — owner, editor, runner, approver, auditor — keep least privilege explicit, and trace logs record the evidence retrieved and the reasoning path of every run.
Agents can use models from Azure AI Foundry (OpenAI GPT, Claude, Llama, Mistral, DeepSeek, Cohere and more) and Google Vertex Model Garden (Gemini and partner models), and a Model Comparison step can pick the best answer across providers. You are not locked into one vendor's roadmap.
Software for building, running and governing AI agents — programs that use a language model to plan and carry out multi-step tasks across business systems. An enterprise AI agent platform adds the controls: access rights, approvals, audit trails, cost limits and testing.
No. Workflows are assembled on a drag-and-drop canvas from nodes for models, knowledge bases, integrations, loops, conditions and human approval. An HTTP Request node covers any system without a ready-made integration.
Through layered controls: safety nodes that check intent, facts, domain boundaries and harmful content; human approval for consequential steps; immutable versions with independent approval before production; role-based access; and full trace logs of every run.
Models available through Azure AI Foundry and Google Vertex Model Garden, including OpenAI GPT, Claude, Gemini, Llama, Mistral, DeepSeek and Cohere, plus embedding models for RAG.
This site needs JavaScript for the full experience. Every page above is reachable without it.