Member-only story
Beyond Agentforce: Agent Fabric, A2A, and Multi-Vendor Orchestration
Orchestrating Agentforce: Building Multi-Agent Systems · Part 5
Everything so far has lived inside Agentforce. One orchestrator, its specialist subagents, all on one platform, all reasoning through Atlas. That’s the right place to start, and for many organizations it’s the whole story. But it’s not where the industry is heading, and it’s not where the hardest problems live.
Here’s the reality most enterprises are walking into: agents aren’t going to come from one vendor. You’ll have Agentforce agents on Salesforce, agents built on Amazon Bedrock, agents on Microsoft Foundry, agents your team wrote from scratch, and agents that arrived with a SaaS product you bought. The average organization already runs a dozen agents, and that number is climbing fast. The question stops being “how do I coordinate my Agentforce subagents?” and becomes “how do I coordinate agents that don’t even share a platform?”
This part is about that world — the protocols and control planes for multi-vendor orchestration. I’ll be honest about what’s production-ready and what’s still maturing, because this is the frontier, and the frontier rewards clear eyes.
(Fast-moving caveat, and it’s the strongest in the series: this is the least settled…









