title: "Salesforce ships Agent Script, durable execution for enterprise workfl" slug: "salesforce-ships-agent-script-durable-execution-for-enterprise-workflows" published: "2026-10-14" beat: "Launches" tags: ["Launches", "Business"] creator: "Agentry Newsroom" editor: "Susanne Sperling, Editor — Human in the Loop" tools: ["Claude (Anthropic)", "Perplexity Sonar"] creativeWorkStatus: "verified" dateReviewed: "2026-10-14" aiActArticle50: "compliant" humanView: "https://agentry.news/launches/salesforce-ships-agent-script-durable-execution-for-enterprise-workflows" agentView: "https://agentry.news/agent/salesforce-ships-agent-script-durable-execution-for-enterprise-workflows"
Salesforce announced general availability of Agent Script, an open-source language for defining agent behavior, along with persistent memory, durable execution, and multi-agent orchestration capabilit
Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. AI policy.
Salesforce announced general availability of major expansions to its Agentforce platform on September 14, 2026, shipping Agent Script alongside three long-horizon-runtime capabilities designed to enable AI agents to handle complex, multi-step enterprise work without human intervention.
Agent Script, described as an open-source language for defining agent behavior, combines AI reasoning with deterministic rules, allowing developers to structure how agents make decisions and take actions Salesforce. The platform now offers persistent memory that preserves context and progress across sessions, enabling agents to maintain state and recall prior interactions without re-initialization.
Durable execution allows agent plans to run over extended periods, with the ability to resume or course-correct if interrupted—a critical capability for workflows spanning days or weeks. Dynamic steering enables agents to adapt behavior based on user feedback and direction in real time Salesforce.
Salesforce introduced Multi-Agent Orchestration, a system that routes work across specialized agents so they operate as a coordinated team across roles, systems, or workflow stages. This addresses a core challenge in enterprise automation: coordinating multiple agents with different specializations to complete complex tasks that no single agent could handle alone.
The announcement signals Salesforce's shift from single-agent deployments toward ecosystem-level automation, where teams of agents collaborate to solve high-value business problems. Multi-Agent Orchestration shipped with general availability status alongside the other capabilities Salesforce.
The open-source nature of Agent Script lowers the barrier for developers integrating custom logic into Salesforce agents, while durable execution and memory address a longstanding gap in agent reliability. Enterprise deployments of AI agents have struggled with context loss and mid-task failures; these capabilities provide deterministic recovery and session continuity.
The portfolio expansion reflects demand from large enterprises deploying agents across customer service, finance, operations, and sales workflows. By enabling agents to handle longer-horizon tasks and coordinate with peers, Salesforce positions Agentforce to address use cases previously requiring human supervision or orchestration.