Next Moca has open-sourced Agent Definition Language (ADL), a vendor-neutral specification that standardizes how AI agents are defined, reviewed, and governed across frameworks and platforms.
Next Moca has open-sourced Agent Definition Language (ADL), a vendor-neutral specification intended to standardize how AI agents are defined, reviewed, and governed across frameworks and platforms. The project is released under the Apache 2.0 license and is positioned as a missing “definition layer” for AI agents, comparable to the role OpenAPI plays for APIs.
ADL provides a declarative format for defining AI agents, including their identity, role, language model setup, tools, permissions, RAG data access, dependencies, and governance metadata like ownership and version history. The aim is to enhance the portability, auditability, and interoperability of agents across various platforms and vendors.
The release addresses a growing fragmentation problem in agent development. Today, agent behavior is often spread across prompts, code, framework-specific configuration files, and undocumented assumptions. This makes it difficult for teams to answer basic questions about an agent’s capabilities, boundaries, and approval status, and complicates security reviews, compliance, and reuse. ADL consolidates agent definitions into a structured, machine-readable format to enhance inspectability and governance.
It is framework-agnostic and focuses on definitions rather than execution, without addressing agent communication, runtime tool invocation, or message transport. ADL is intended to complement existing technologies like A2A, MCP, OpenAPI, and workflow engines.
Announcing the open-source release, Next Moca founder Kiran Kashalkar described ADL as “Think OpenAPI (Swagger) for agents,” adding that it provides “a single, declarative spec that says what an agent is, what tools it can call, what data it can touch, and how it is configured.” Kashalkar emphasized portability, auditability, and vendor neutrality as core design goals.
According to Next Moca, ADL is aimed at teams building production AI systems, where agents increasingly operate as autonomous components with access to tools, data, and external systems. The company argues that a standardized definition layer enables clearer planning, consistent validation in CI pipelines, explicit comparison of agent capabilities, and software-style lifecycle management through versioning and rollback.
The project includes a published JSON Schema, example agent definitions, validation tools, and documentation covering governance and contribution processes. Developers can define an agent once and validate it locally, then share the same definition across security, platform, and compliance teams.
Next Moca describes ADL as an early-stage standard and is inviting community feedback and contributions to shape its evolution. The company states that open-sourcing the specification is intended to encourage broad adoption, neutral governance, and the development of an ecosystem of editors, validators, registries, and testing tools around a shared format.
The ADL repository and documentation are available on GitHub, with contribution guidelines and a public roadmap outlining planned next steps.
The Agent Definition Language specification represents a significant step toward standardizing AI agent development in the same way OpenAPI standardized REST API development. By providing a vendor-neutral, declarative format for defining agents, ADL addresses the fragmentation that currently exists across different frameworks and platforms.
This standardization could have far-reaching implications for enterprise AI adoption. When agents are defined using a common specification, organizations can more easily audit their capabilities, ensure compliance with security policies, and migrate between different agent platforms without rewriting definitions. The governance metadata support is particularly valuable for regulated industries where tracking ownership, version history, and approval status is critical.
The comparison to OpenAPI is apt - just as OpenAPI enabled the creation of tools like Swagger UI, Postman, and API gateways that work across different backend technologies, ADL could enable similar tooling for agent management. This might include agent registries, validation tools, testing frameworks, and even marketplaces where agents can be discovered and deployed.
However, the success of ADL will depend on community adoption. The specification is still in early stages, and its evolution will be shaped by contributions from developers and organizations building AI agents. The framework-agnostic approach is smart, as it allows ADL to complement rather than compete with existing agent frameworks and communication protocols.
For developers working with AI agents today, ADL offers a path toward more maintainable, auditable, and portable agent definitions. Instead of scattering agent configuration across multiple files and undocumented assumptions, teams can consolidate everything into a single, validated specification that can be version-controlled and reviewed like any other piece of software.
As AI agents become more prevalent in production systems, having a standardized way to define and govern them becomes increasingly important. ADL represents an important step in that direction, and its open-source nature ensures that the specification can evolve based on real-world usage rather than being dictated by a single vendor.
The timing of this release is also noteworthy, coming at a point when organizations are moving beyond experimental AI projects to production deployments where governance, security, and interoperability become critical concerns. ADL provides the tooling and standardization needed to make this transition more manageable.
For teams building AI agents, the question is no longer whether to adopt a standardized definition format, but rather which format to adopt. With ADL now available as an open-source specification, it presents a compelling option that prioritizes portability, auditability, and vendor neutrality - qualities that will be essential as the AI agent ecosystem continues to mature.
Comments
Please log in or register to join the discussion