
Decawork is an agent control plane training company-specific models to approve, block, or hold AI agent actions in real time.
decawork.aiCrawled Oct 15, 2026
Decawork trains small, company-specific models to approve AI actions in real time, using company policies, task context, and past actions.
The page describes information Decawork collects through its website, how it is used and shared, how long it is retained, and available choices.
Decawork's marketing website terms cover site use, product information, ownership, third-party services, warranties, and liability.
Product Tapas' August 27, 2026 edition features Decawork among the week's new tools, describing it as a control panel for a company's internal AI agents and tools. The product lets IT set permissions, monitor usage, stop runaway agents, and audit the AI workforce's activity. The edition also rounds up product news involving Anthropic, Apple, and Uber, but the supplied text gives no further details.
Aug 2026Product TapasNews article
Decawork co-founder Sarthak Aggarwal says enterprise AI agents need an IT-style department focused on employment readiness, not just more capable models. He identifies identity, access, delegation, audit, and revocation as essential controls and proposes privilege separation for agents. The article cites the EchoLeak and Replit incidents as evidence that instructions alone are not controls.
Aug 2026BigGo FinanceNews article
StartupKit probes careers pages and security.txt intake channels across the 80 newest Launch YC companies. Its August 13, 2026 probe highlights Decawork, an agent-control-plane security company launching August 12 with 135 votes, because its only security contact routes to a founder alias. The article says launch-week companies need managed intake channels rather than founder inboxes.
Aug 2026StartupKit BlogNews article
Decawork AI launches a platform that helps IT teams deploy, govern, and maintain internal AI agents. Each agent receives its own identity and scoped credentials, with IT approval, monitoring, and spending controls applied before deployment. The platform sits between internally built agents and enterprise systems as an operational control layer.
Aug 2026DealroomNews article