xpander.ai Launches Vendor Neutral AI Agent Control Plane, Raises $7.5 Million
xpander.ai launches a vendor-neutral AI agent control plane and Omni, raises $7.5 million, and targets enterprise governance, portability and agent sprawl.
Summary
On August 17, 2026, xpander.ai, founded by three former AWS principal engineers and led by CEO David Twizer, made its enterprise agent platform and Omni generally available. Its vendor neutral Universal Harness and control plane coordinate execution, permissions, identity, observability, memory, auditability and lifecycle management across models, frameworks and clouds. The launch includes a $7.5 million seed round led by Pico Venture Partners, with Emerge Ventures, Samsung Next and SeedIL. Gartner expects an average global Fortune 500 company to exceed 150,000 agents by 2028, from fewer than 15 in 2025, while only 13% report adequate governance.
Deployment spans xpander’s cloud, Kubernetes, AWS, Google Cloud, Microsoft Azure, private VPCs, on premises and air gapped systems. It supports LangChain, Strands and Agno agents; proprietary, open weight and fine tuned models; and REST, Python SDK and Model Context Protocol integrations for Claude Desktop and Cursor. Controls include approvals, named identities, user attribution, audit and spending logs, vault injected credentials, SSO, OIDC and private model gateways; xpander claims SOC 2 Type II certification and GDPR compliance. Multiplayer AI preserves permission scoped workflows across Slack, Teams, ChatGPT, Claude and xpander. Omni creates shareable Agentic Applications; xpander reports 90.9% on GAIA using Opus, Sonnet, Kimi and GPT, without independent validation.
Team pricing permits unlimited seats, agents and workflows: one credit equals $0.01, each agent wake or tool or API call costs one, and model tokens cost extra; $100 buys 10,000 credits and new accounts receive 1,000. Self hosted Enterprise is an undisclosed annual license starting at 50 agents, plus infrastructure and possible model costs. LangSmith, CrewAI, Temporal, OpenAI Frontier, Google Gemini Enterprise Agent Platform and Amazon Bedrock AgentCore overlap. xpander must prove its neutrality and collaboration justify another layer; its proprietary harness may relocate lock in because exit portability is undocumented.
Positives
- $7.5 million from Pico Venture Partners, Emerge Ventures, Samsung Next and SeedIL gives xpander funding to expand its enterprise platform.
- AWS, Google Cloud, Microsoft Azure, Kubernetes, private VPC and air gapped deployments give customers broad infrastructure choices.
- Vault injected credentials keep secrets from models, while named identities, approvals and logs strengthen accountability.
- Unlimited Team seats, agents and workflows remove seat charges, with 1,000 free credits for new accounts.
- 90.9% on GAIA, using Opus, Sonnet, Kimi and GPT, supports xpander’s multi-model harness thesis, although the result is company reported.
Risks & concerns
- Only 13% of organizations report adequate governance as Fortune 500 agent counts are forecast to exceed 150,000 on average by 2028.
- xpander’s proprietary Universal Harness could replace model or cloud dependence with control plane lock in.
- Public documentation does not explain whether configurations and operational state remain portable after an enterprise license ends.
- Enterprise self hosting requires an undisclosed annual license beginning at 50 agents, preventing public total cost calculations.
- LangSmith, CrewAI, Temporal, OpenAI, Google and Amazon already provide overlapping portability, governance or execution capabilities.