Salesforce Just Made Cross-Platform Agent Memory Official: What Agent Context Engine Means for Your Scheduled Agents

Salesforce Just Made Cross-Platform Agent Memory Official: What Agent Context Engine Means for Your Scheduled Agents

At Dreamforce, Salesforce announced something bigger than a product update. It announced Agent Context Engine (ACE), a shared context layer for AI agents — and in doing so, validated an idea that automation operators have lived with for years: an agent's memory should not stop at the platform boundary.

Here is what ACE is, why Salesforce built it, and what it means for your scheduled agents running outside the Salesforce universe.

What Agent Context Engine actually is

ACE is a shared context layer that works across agents and across platforms. According to Salesforce's announcement, any agent on a supported platform can pull context from it — agents built on Agentforce, Google ADK, AWS Bedrock, LangGraph, or the OpenAI SDK. Agents built outside Agentforce connect through SDK code or MCP tools, and every agent receives the same governed context, filtered to what it is permitted to see.

Under the hood, ACE sits on Salesforce's Data 360 customer data platform: the real-time pipeline, zero-copy access to data where it lives, identity resolution, tagging, and governance. On top of that foundation, ACE adds a knowledge-graph and traversal engine that assembles context using GraphRAG. When a request arrives, ACE translates the incoming prompt into the minimum data needed for the maximum result, walking the graph from the resolved customer and traversing only to the facts that are relevant and permitted, within the business context you defined.

What the agent receives is a scoped payload assembled for that one request: the resolved customer profile, the relevant memories, the real-time signals about what the customer is doing right now, and the slice of the knowledge graph relevant to the scenario. Every element is filtered to what this agent is permitted to see, which Salesforce says keeps the interaction responsive while holding inference cost down.

Two details matter most for operators:

Context compounds. ACE mines the exhaust every agent produces as it works — the raw conversation and the agent's reasoning steps — for facts, preferences, and learnings, and persists them with provenance and governance. One agent's write becomes the next agent's read. Salesforce is explicit that this can require human confirmation before other agents rely on it, because an unchecked memory means one agent's guess becomes another agent's "fact."

Context is governed. Assembly applies the entitlements you already enforce: object, field, and row-level security, sharing rules, data classification. Sensitive content is masked or withheld before it reaches the agent. What an agent can read and what it can write are both governed, and every memory is traceable to its source.

ACE is in preview now; Salesforce says you can start using it today with agent registration and configuration.

Why Salesforce built it

Salesforce opens the announcement with a story: an account executive opens a $5M expansion that the CRM marks "Commit" and asks an agent for the next best action. Working from the CRM alone, the agent drafts a confident close email. Meanwhile the enterprise knows product usage has fallen 27%, there is an open billing dispute, an unsigned compliance rider, and a customer success agent who — after an outage — proposed holding renewal pricing flat and waiving the overages. With shared, assembled, governed context, the agent surfaces all of that and recommends a service update to the exec sponsor instead of firing off the close.

The diagnosis, in Salesforce's own words: "The gap is not data. It is context." Every company already owns identity systems, business processes, data platforms, governance, and knowledge bases in some form — but the pieces are fragmented and never assembled. Memory, where it exists at all, belongs to a single agent rather than every agent and person in the company. And while agent platforms have started to provide memory and learning, they stop at the platform boundary. When a customer moves between agents, channels, and people, the next interaction loses what the enterprise already knows.

That last admission is the important one. Salesforce just said, in public, that per-platform agent memory is a dead end.

What this means for your scheduled agents

You run the same problem at operator scale. A lead-scoring agent in n8n learns that a target account just changed its pricing page; the next morning, a follow-up agent in Make re-researches the same account from scratch. A weekly reporting agent in Zapier rediscovers the same broken integration every single week. Each automation wakes up blind, re-briefs itself from scratch, and whatever one run figured out dies with it — unless the memory lives somewhere outside all of them.

ACE is the enterprise answer to exactly this problem. It is built for companies on the Salesforce stack, with Data 360 as the foundation, and it is aimed at the full weight of enterprise governance: entitlements, masking, provenance, human confirmation. If that describes your world, the path is agent registration and the preview signup.

But if your agents live in n8n, Make, Zapier, and a few Python scripts on a cron, you need the same pattern without the enterprise stack: a memory layer that sits outside every platform, connects over an open protocol, and hands every run the same continuous context. The layer, not the platform, is what remembers.

That is the pattern Vilix AI is built for: cloud-hosted with zero infrastructure to manage, the same memory available to every agent everywhere over MCP, full conversation history rather than extracted facts alone — so the reasoning behind a decision can be revisited, not just its outcome. It comes with a free plan that is free forever, and a 7-day Pro trial with no credit card required. Your data stays yours: export everything or delete it anytime. One memory, every scheduled agent, every tool, every run.

The bottom line

Salesforce did not just announce a feature. It announced a category position: shared, governed, cross-platform agent memory is where the industry is going, and the per-platform memory of today is the legacy to replace. The giants are building it for the enterprise. Your scheduled agents need it now — and they don't need an enterprise data platform to get it.