We open-sourced our agent platform's governance layer: hexagonal architecture and governed execution

Описание и применение

SOIT предоставляет открытый агентский рантайм с встроенным управлением, позволяя компаниям безопасно запускать ИИ‑агентов в продакшене, контролируя доступ, секреты, исходящий трафик, аудит и расходы.

Задача

Teams can build AI agents but struggle to trust them in production due to lack of governance, credential management, egress control, auditability, and cost visibility.

Результат

SOIT enables teams to run AI agents in production with full governance, auditability, and cost visibility, while remaining open‑source and vendor‑agnostic.

Рабочий процесс

• 1. Define agents and workflows in the visual editor with versioning.
• 2. Configure approved models, tools, and knowledge bases via per‑version allowlists.
• 3. Store credentials in Vault and inject them through the gateway.
• 4. Set egress policies to restrict outbound HTTP calls.
• 5. Deploy SOIT with a single Docker‑Compose command.
• 6. Run agents; each step is recorded in the runtime ledger.
• 7. Monitor live metrics, cost, and failures in the workspace console.
• 8. Replay or audit runs using the trace timeline.

Доказательства

Every agent run is a governed run. Chat turn, agent loop, or workflow run — everything flows through one runtime ledger (Run / RunStep / Trace), and the same controls apply on every path:

Источник

The platform self‑hosts with one command: git clone … docker compose … up -d

Источник