# OutSystems Agent Workbench on the GAMMA-Ray Spectrum: A4 · R3 · C2 · O2

> GAMMA-Ray v0.0.166 (beta, for review; every rating is experimental). The GAMMA-Ray studio reads OutSystems Agent Workbench from its public documentation: agents call the actions a builder gives them, and with workflows and A2A they chain, delegate and stop for a person.
> Page: https://gamma-ray.sgit.ai/reports/outsystems/article.html · Source: data/reports/outsystems.json


PrompTugal.news draft · from the GAMMA-Ray report · beta · 2026-11-10

The GAMMA-Ray studio reads OutSystems Agent Workbench from its public documentation: agents call the actions a builder gives them, and with workflows and A2A they chain, delegate and stop for a person.

OutSystems, founded in 2001 by Paulo Rosado in Lisbon, Portugal, sells a low-code development platform. Its Agent Workbench, generally available since 30 September 2025, builds and runs AI agents on OutSystems Developer Cloud. We read it only from public sources, captured and hashed on 10 November 2026, and the published rules computed the card.

**Autonomy 4, Orchestrates (available).** Agents choose and call actions in other systems. In a workflow, several agents pass work along, and a person approves or rejects output at high-stakes steps. TeamWork's agent classifies and routes support tickets.

**Reach 3, Open protocol (available).** With an A2A connection, an agent delegates tasks to agents outside ODC over the Agent2Agent protocol, and each agent reaches only the actions added to it.

**Control 2, Configurable (available).** A human activity in a workflow holds the agent until an assigned person acts. A public record of agents monitored in operation would reach level 3.

**Openness 2, Documented.** The docs list the model providers, the grounding data and where guardrails do not apply. Published evaluation results would reach level 3.

The card is beta and OutSystems has not seen it. It can claim the card and add evidence; corrections are dated.

From public sources only; OutSystems has not seen it. The levels come from [the report](./index.md) and its cards, which are corrected in place. Quote the card by linking to it.

## O Agent Workbench da OutSystems no GAMMA-Ray Spectrum: A4 · R3 · C2 · O2

O estúdio GAMMA-Ray lê o Agent Workbench da OutSystems a partir da sua documentação pública: os agentes chamam as ações que lhes são dadas e, com workflows e A2A, encadeiam-se, delegam e param para uma pessoa.

A OutSystems, fundada em 2001 por Paulo Rosado em Lisboa, vende uma plataforma de desenvolvimento low-code. O Agent Workbench, disponível desde 30 de setembro de 2025, cria e executa agentes de IA. Lemos apenas fontes públicas, capturadas e com hash a 10 de novembro de 2026.

**Autonomia 4 (opcional).** Num workflow, vários agentes passam o trabalho entre si e uma pessoa aprova ou rejeita nos passos de maior risco. **Alcance 3 (opcional).** Com uma ligação A2A, um agente delega tarefas a agentes fora da plataforma. **Controlo 2 (opcional).** Uma atividade humana retém o agente até a pessoa designada agir.

**Abertura 2.** A documentação lista os fornecedores de modelos e diz onde as guardrails não se aplicam. O cartão é beta e a OutSystems ainda não o viu; pode reclamá-lo e acrescentar evidência, e as correções são datadas.