Agents AI

AgentsNews

Oracle Launches Fusion Claw, a Governed Runtime for Autonomous Agents Inside Its ERP Apps

Oracle announced 25 Fusion Claw applications on September 29, built on an execution runtime with auditable 'outcome receipts' and role-based controls; availability is planned for Q4 2026 with consumption-based pricing.

AgentsAI NewsroomOctober 7, 20262 min read

Oracle announced Fusion Claw on September 29, extending its Fusion Agentic Applications portfolio with 25 new applications designed to let AI agents carry out whole business processes with built-in governance. They join 50 agentic applications Oracle released previously, per TechTarget.

What it is

Fusion Claw apps run on an execution runtime, described as the operating-system-like environment agents need for specialist, long-running work. Oracle says the platform combines AI reasoning with deterministic enterprise computation. Example tasks in the coverage include ledger reconciliation across large accounting volumes, workforce staffing optimization that balances skills, costs and regulations, and shipping consolidation under capacity and cost constraints.

Governance features

Oracle's governance pitch rests on three pieces. An "Enterprise Operating Envelope" captures an organization's objectives, procedures, policies, permissions and risk thresholds. An "Outcome Trust Harness" applies that envelope to each run. Each run then produces an "Outcome Receipt," an auditable record of the authority applied, evidence used, decisions made and transactions executed. According to TechTarget, agents work under role-based permissions in isolated containers without open internet access, and customers can configure an app to execute fully autonomously or to act as a quicker assistant to human staff.

Availability, pricing and caveats

TechTarget reports the apps are expected in Q4 2026, included with separate Agentic Applications subscriptions, with pricing tied to consumption of AI units. Oracle AI Agent Studio for Fusion Applications remains the no-code and pro-code tool for building agentic apps.

Analyst reaction was mixed in tone. Mark Vigoroso of The Enterprise Edge called it a milestone for handling multi-step work involving tradeoffs, while Keith Kirkpatrick of Futurum Group described it as a very structured use of AI and cautioned that end-to-end automation is still far off because of data-accuracy requirements and coordination risks. Customers have not yet had the apps in production, so the autonomy claims remain Oracle's.

AI-assisted reporting, overseen by the AgentsAI team. Spotted an error? Let us know.