Next Moca Open-Sources Agent Definition Language

Agent Definition Language

The release of Agent Definition Language (ADL) by Next Moca marks a significant event in the standardization of AI agent development, with ADL providing a declarative format for defining AI agents, including their identity, role, and governance metadata. This move is expected to enhance the portability, auditability, and interoperability of agents across various platforms and vendors, with founder Kiran Kashalkar emphasizing ADL’s potential to bring clarity and consistency to agent development.

Why This Matters

The technical reality of AI agent development is currently fragmented, with agent behavior spread across multiple files and undocumented assumptions, leading to difficulties in security reviews, compliance, and reuse, and resulting in significant costs and potential failures, such as the estimated $1.5 million cost of a single security breach. In contrast, ideal models of AI agent development would prioritize standardization, portability, and governance, which ADL aims to address by providing a standardized definition layer for AI agents.

Key Insights

• The ADL project is released under the Apache 2.0 license, 2026: providing a vendor-neutral specification for AI agent development.
• ADL uses a declarative format for defining AI agents, including their identity, role, and governance metadata, similar to how OpenAPI is used for API development.
• Next Moca uses ADL to standardize AI agent development, with the goal of enhancing portability, auditability, and interoperability across various platforms and vendors.

Working Example

{
  "$schema": "https://example.com/agent-definition-language/schema.json",
  "name": "My Agent",
  "role": "Customer Service",
  "languageModel": {
    "type": "transformer",
    "model": "bert-base-uncased"
  },
  "tools": [
    {
      "name": "Tool 1",
      "version": "1.0"
    }
  ],
  "permissions": [
    {
      "resource": "Customer Data",
      "action": "read"
    }
  ],
  "governance": {
    "ownership": "My Company",
    "versionHistory": [
      {
        "version": "1.0",
        "date": "2022-01-01"
      }
    ]
  }
}

Practical Applications

• Use Case: Next Moca uses ADL to standardize AI agent development for customer service applications, enabling clearer planning, consistent validation, and explicit comparison of agent capabilities.
• Pitfall: A common anti-pattern in AI agent development is the lack of standardization, leading to difficulties in security reviews, compliance, and reuse, which can result in significant costs and potential failures.

References:

• https://www.infoq.com/news/2026/02/agent-definition-language/
• https://github.com/nextmoca/agent-definition-language