What is the AI agent standards war? The AI agent standards war is the 2026 fight between rival open specifications that decide how AI agents connect to, discover, and use the software where commercial real estate data lives. On one side sits Anthropic's Model Context Protocol (MCP), the de facto standard for tool access since late 2024. On the other, a coalition led by Google and Microsoft is backing a newer specification called ARD (Agentic Resource Discovery). For CRE investors wiring AI agents into their tech stack, the outcome shapes integration cost, vendor lock-in, and how fast agents can work across platforms like Yardi, RealPage, and CoStar. For the broader toolkit, see our complete guide on AI tools for real estate investors.
Key Takeaways
- The AI agent standards war pits Anthropic's MCP, the incumbent tool-access standard, against ARD (Agentic Resource Discovery), a June 17, 2026 spec backed by Google, Microsoft, and nine other firms.
- MCP, A2A, and ARD are complementary layers, not direct rivals: MCP connects agents to tools, A2A connects agents to each other, and ARD lets agents discover which tools exist.
- For CRE firms, the standard your vendors support determines integration cost and lock-in risk across systems like Yardi, RealPage, MRI, Dealpath, and Argus.
- Anthropic and OpenAI are absent from ARD's backer list, yet all sides still cooperate inside the Linux Foundation's Agentic AI Foundation.
- The practical move for CRE buyers is to favor open-standard support and architect for flexibility, not to bet on a single vendor's proprietary integration.
The AI Agent Standards War Explained
The AI agent standards war is a contest over the invisible plumbing that lets AI agents plug into company data and tools. In July 2026, The Information reported that Google, Microsoft, Salesforce, Snowflake, and ServiceNow had agreed to back a shared technical standard for connecting agents to business software, a move widely read as aimed at Anthropic and OpenAI. Anthropic's MCP had quietly become the default over the prior 18 months, and the largest enterprise software vendors were reluctant to keep building on a competitor's foundation.
The specification at the center of the coalition is ARD, or Agentic Resource Discovery, announced June 17, 2026 by eleven companies: Cisco, Databricks, GitHub, GoDaddy, Google, Hugging Face, Microsoft, NVIDIA, Salesforce, ServiceNow, and Snowflake. It uses an open, Apache 2.0 license, so any vendor can implement it and no single backer can lock it down. Notably, the two companies most associated with frontier AI models, Anthropic and OpenAI, were nowhere on the initial supporter list.
MCP, A2A, and ARD: Three Layers, Not Three Rivals
Despite the war framing, these standards mostly solve different problems, and understanding the three-layer stack keeps CRE technology decisions grounded. Each answers a distinct question an AI agent asks when it goes to work on a deal or a portfolio.
- MCP (Model Context Protocol): Created by Anthropic in November 2024, MCP is the "USB-C for AI," connecting an agent to a specific tool, database, or API. By April 2026 its Python SDK alone crossed 164 million monthly downloads, and every major platform, including ChatGPT, Gemini, Claude, and Microsoft Copilot, supports it.
- A2A (Agent to Agent): Introduced by Google in April 2025 and now governed by the Linux Foundation, A2A lets one agent hand work to another. It reached a stable version 1.0 in April 2026.
- ARD (Agentic Resource Discovery): ARD is the missing discovery layer. It lets an agent look up which tools and agents exist for a task instead of being told exactly where each one lives, using an ai-catalog.json file published on an organization's own domain.
In plain terms: MCP is how an agent talks to your rent-roll parser, A2A is how that agent hands the output to an underwriting agent, and ARD is how either agent finds those tools in the first place. We covered the vendor side of this shift when a proptech platform shipped its first MCP server, but the standards war plays out one layer above any single vendor.
Why the Standards War Matters for CRE Investors
For CRE investors, the standards war matters because it decides how much it costs to connect AI agents to the systems that run your portfolio. A mid-sized owner-operator might run Yardi or MRI for accounting, RealPage or AppFolio for property management, Dealpath for pipeline, and Argus for valuation, with CoStar for market data. Every AI agent you deploy has to reach into that stack. When those vendors support a common open standard, an agent can read a T12, compute NOI (gross revenue minus operating expenses), and flag any asset with a DSCR below 1.25x without a custom integration for each system.
The alternative is expensive. Proprietary, one-off integrations are a big reason so many CRE AI pilots stall. Industry data shows that while 92% of corporate occupiers have initiated AI programs, only 5% report achieving most of their AI goals, and integration friction is a major culprit. A settled, open standard shrinks the cost of connecting an agent from weeks of engineering to a configuration step, moving it from a demo to a tool that touches real cap rate and IRR analysis. The AI in real estate market is forecast to reach $1.3 trillion by 2030 at a 33.9% CAGR, and most of that value depends on agents that can actually reach the data.
What CRE Firms Should Do Now
The right response to a standards war is not to crown a winner today but to architect for flexibility, because the backers themselves are hedging. Here is a practical checklist for CRE technology buyers.
- Ask vendors which standards they support. Any proptech or AI vendor you evaluate in 2026 should be able to say whether it supports MCP, and increasingly whether it is ARD-discoverable. A vendor offering only a closed, proprietary integration is a lock-in risk.
- Favor open standards over bespoke connectors. Open, Apache 2.0 standards like ARD lower switching costs, which protects your firm if you change property management or valuation platforms later.
- Separate the standard from the model. Which large language model you use, whether Claude, GPT, or Gemini, is a separate decision from which connectivity standard your tools speak. Keeping them decoupled preserves optionality on both.
- Treat governance as part of the stack. Open discovery makes agents more capable and widens the attack surface, so pair standards adoption with the controls in our guide to AI agent governance for CRE firms.
For personalized guidance on building an agent-ready CRE tech stack, connect with The AI Consulting Network, which helps owners and operators design their tech stack for interoperability.
The Cooperation Paradox
The strangest feature of the 2026 standards war is that the combatants are also allies. Google, Microsoft, Salesforce, Snowflake, OpenAI, and Anthropic are all members of the Linux Foundation's Agentic AI Foundation (AAIF), which governs both MCP and A2A across more than 140 member organizations, including JPMorgan Chase and American Express. They cooperate on shared open standards inside the foundation while competing in the market over whose plumbing developers actually use.
For CRE investors, that paradox is reassuring. Because the major standards are open and governed by a neutral body, the risk of a single walled garden locking up your agent stack is lower than the headlines suggest. The rational posture is to ride the open standards, watch which one your critical vendors adopt, and avoid proprietary dead ends. This mirrors the cost discipline in the Cisco model-routing playbook for CRE firms: the winners treat AI infrastructure as a portfolio of swappable parts, not a single bet. If you are ready to build that kind of resilient AI stack, The AI Consulting Network specializes in exactly this.
Frequently Asked Questions
Q: What is the difference between MCP and ARD?
A: MCP (Model Context Protocol) connects an AI agent to a specific tool or data source, like linking an agent to your Yardi database. ARD (Agentic Resource Discovery) is a discovery layer that lets an agent find which tools and agents exist in the first place. They are complementary, not direct substitutes.
Q: Does the AI agent standards war affect small CRE firms?
A: Yes, but mostly as buyers rather than builders. Small and mid-sized CRE firms benefit when their vendors adopt open standards, because it lowers integration cost and lock-in. The practical action is to ask vendors which standards they support before signing.
Q: Should I wait for a winner before deploying AI agents?
A: No. MCP, A2A, and ARD are open standards designed to interoperate, and most major enterprise platforms already support more than one. Deploy now on vendors that back open standards, and architect for flexibility rather than betting everything on one specification.
Q: Is Anthropic losing the AI agent standards war?
A: Not clearly. MCP has an 18-month head start, more than 164 million monthly SDK downloads, and near-universal platform support. The Google and Microsoft coalition has enormous enterprise reach, but committee-designed standards move slowly. Most analysts advise hedging across both rather than declaring a winner.