Fortune 500 companies will average 150,000 AI agents by 2028, up from fewer than 15 in 2025, according to Gartner. Only 13% of organizations believe they have adequate governance in place. xpander.ai, founded by three former AWS principal engineers, launched its enterprise agent platform today with $7.5 million in seed funding led by Pico Venture Partners, with Samsung Next and Emerge Ventures participating. The platform is a vendor-neutral control plane for building, running, and governing agents across different models, frameworks, and infrastructure environments.
The core product is a Universal Harness: a runtime that executes agents as portable workloads across AWS, Google Cloud, Azure, private VPCs, and air-gapped on-premises environments. Developers get a REST API, a Python SDK, and Model Context Protocol support compatible with Claude Desktop and Cursor. xpander supports LangChain, Strands, and Agno frameworks, plus proprietary, open-weight, and fine-tuned models. CEO David Twizer frames the pitch around avoiding vendor lock-in, drawing on his seven years at AWS watching enterprises repeat that mistake with cloud infrastructure. The catch: xpander's own control plane and proprietary harness become the new dependency. The company's public documentation does not yet explain how portable customer configurations are if they terminate their enterprise license.
The competitive landscape undercuts any simple neutrality argument. LangChain's LangSmith Deployment already supports air-gapped Kubernetes deployments. CrewAI offers SSO, role-based access, and customer-owned infrastructure. Temporal handles durable execution for long-running workflows. OpenAI and Google are both expanding upward into this governance and runtime layer. The article is worth reading in full for its breakdown of where each competitor draws the line between managed control planes and customer-owned infrastructure, and for its analysis of how xpander's lock-in trade-off compares to the problem it claims to solve.
[READ ORIGINAL →]