iAureo Agents Platform

iAureo Agents Platform gives organisations a common place to create, install, run and supervise AI agents. It is independent of any specific business application: an ERP, document system or customer-facing service can integrate through an API, SDK, webhooks or supported connectors.

The agent is the central unit. Each published version defines a purpose, the capabilities it needs, the tools it may use and its operating limits. A company then installs that version with its own model configuration, knowledge sources and permissions. Publishing a definition and allowing it to act are explicit administrative decisions.

The platform is in production, with continued development around integrations, evaluation and user experience. It complements iAureo Platform's shared document and communication services while remaining a separate product with its own execution and governance model.

Calling an AI model does not settle who may use an agent, which business actions it may perform or what should happen when a process needs approval. A production system also needs to preserve work across interruptions, record the version and configuration that ran, and make usage costs visible.

The challenge was to provide those capabilities across companies and applications without giving a model unrestricted access to business data. Model providers, tools, documents and integrations needed to remain configurable resources with explicit boundaries.

Agents with published, traceable versions

Agents can be created directly, from a template or through a conversational builder. All paths lead to validation and explicit publication. A published version is immutable; the installation in each company determines the version, model profiles, resources and limits available to it. The execution records a snapshot of that configuration before invoking a model.

Tools that control real actions

Models act through registered tools with declared inputs, risk, time limits and approval requirements. A tool can be called by an agent, a workflow or an authorised application through the same execution mechanism. High-risk actions require human approval, and the model does not receive unrestricted SQL access. Document-processing capabilities run in separate services and return references to generated files.

Workflows that can wait and resume

The workflow engine persists processes and their progress. Steps can run agents or tools, evaluate conditions, wait for approval or an external event, and schedule delays. Retry policies, parallel branches and recovery support work that extends beyond a single request. The visual Workflow Studio brings together the process definition, input contract and authorised applications.

Company knowledge and model choice

The data2go knowledge library organises documents, versions and publication states. Agents retrieve authorised published material. Model profiles separate an agent's needs from a specific provider deployment; company configuration determines available providers, models, fallbacks and budgets. Usage records make the model, tokens and associated cost visible for each agent invocation.

Supervision at the point of execution

The administration console shows execution timelines, tool calls, approvals, results and consumption. Declared guardrails can flag, redact, escalate or block content. Approvals can pause a sensitive action or hold an output for review. Company and application permissions determine which agents, tools and workflows an integration can reach.

Integration into existing products

The platform exposes a client API, .NET SDKs, signed completion callbacks and MCP integration. The implementation also includes messaging and voice services for configured channels. A business application can start work, follow progress and retrieve the result without knowing the internal database or worker structure.

A production platform with a defined next stage

.NET 10, Blazor, EF Core and SQL Server underpin separate API, administration and background processes, deployed on Linux with Docker. Development continues around the embedded experience, cost consolidation and evaluation. Direct delegation from one agent to another remains planned; current workflows coordinate their declared steps explicitly.

Learn about our AI and automation services and the separate iAureo shared-services platform.