The emergence of agents.json as a new specification for enabling AI agents to interact with APIs has sparked significant discussion within the developer community. Created by Wildcard AI, this open specification aims to bridge the gap between traditional API structures and the needs of Large Language Models (LLMs), but community feedback reveals both enthusiasm and concerns about its approach, licensing, and potential adoption.
 |
|---|
| Founders of Wildcard AI discussing their new specification for AI agent interaction |
Stateless vs. Stateful Protocols: The MCP Comparison
One of the most prominent discussions surrounding agents.json centers on its stateless approach compared to the Model Context Protocol (MCP). While MCP maintains a 1:1 connection between client and server with shared context, agents.json deliberately takes a stateless path more familiar to traditional API development. This fundamental difference has divided the community on which approach better serves the future of AI agent development.
MCP is great for the stateful systems, where shared context is a benefit, but this is a rarity. Developers generally write clients to use APIs in a stateless way, and we want to help this majority of users.
The Wildcard AI team maintains that the two protocols aren't mutually exclusive, suggesting that agents.json fills a gap for developers who prefer managing state within their applications rather than relying on external entities. Some community members speculate that both approaches might coexist, serving different use cases in the emerging AI agent ecosystem.
Licensing Concerns Threaten Adoption
Perhaps the most significant barrier to agents.json adoption identified in community discussions is its licensing structure. While the specification itself is licensed under Apache 2.0, the Python implementation package uses the more restrictive AGPL license. This has prompted concerns about commercial viability and integration potential.
Several developers have questioned how the AGPL-licensed package could be incorporated into commercial products, with some describing it as dead on arrival due to licensing restrictions. The Wildcard AI team has acknowledged these concerns, explaining that they primarily want to prevent large cloud providers from turning their work into a proxy service, while still allowing integration with open-source frameworks like LangChain or CrewAI.
This licensing debate highlights the delicate balance open-source AI projects must strike between protecting their work and fostering widespread adoption.
Simplicity vs. OpenAPI Compatibility
The community appears divided on whether agents.json should maintain strict compatibility with OpenAPI or prioritize simplicity. Some developers appreciate the OpenAPI foundation, noting that many API documentation platforms can export OpenAPI specifications. Others question whether this compatibility adds unnecessary complexity, pointing out that OpenAPI hasn't achieved universal adoption.
The Wildcard AI team has indicated that while they're currently building on OpenAPI, they're designing agents.json to be extensible beyond REST APIs, with plans to support GraphQL, gRPC, and internal SDKs. This suggests a future where agents.json might evolve beyond its OpenAPI origins while maintaining backward compatibility.
Documentation and Usability Challenges
Several community members have highlighted difficulties in understanding and implementing agents.json due to documentation issues. Specific complaints include challenges finding example files in the registry and a lack of clear, immediately accessible examples in the documentation.
In response, the Wildcard AI team has added download buttons to their registry and acknowledged the need for improved documentation. They've also mentioned plans to develop tools that would make creating agents.json files easier, including a validator and an interactive builder.
 |
|---|
| A live demo showcasing the Resend API's interface, illustrating usability challenges with current APIs |
Business Model Questions
As a Y Combinator-backed company, Wildcard AI faces questions about how they plan to monetize a protocol. The team has indicated their primary revenue strategy involves charging API providers for white-glove implementation of the standard, rather than charging end developers who consume the agents.json files.
This approach has prompted some skepticism about whether larger companies will adopt a standard created by a startup, though the team mentions that companies like Resend and Alpaca have already shown interest in implementing the protocol.
In conclusion, agents.json represents an interesting approach to solving the challenges of enabling AI agents to interact effectively with APIs. While the community recognizes its potential value, questions about licensing, compatibility, documentation, and business viability will likely determine whether it gains widespread adoption or remains a niche solution. As AI agent capabilities continue to evolve rapidly, the need for standardized interaction protocols becomes increasingly important, suggesting that either agents.json or a similar solution will eventually emerge as the standard.
Reference: Translate OpenAPI into LLM Tools with agents.json