An Internet of AI Agents? Coral Protocol Introduces Coral v1: An MCP-Native Runtime and Registry for Cross-Framework AI Agents

Coral Protocol has launched Coral v1, an agent stack designed to standardize the discovery, composition, and operation of AI agents across various frameworks. This release focuses on a Model Context Protocol (MCP)-based runtime, known as Coral Server, which facilitates threaded, mention-addressed agent-to-agent messaging. Additionally, it includes a developer workflow (CLI + Studio) for orchestration and observability, along with a public registry for agent discovery. Coral plans to introduce pay-per-usage payouts on Solana, which are not yet generally available.

What Coral v1 Actually Ships

For the first time, users can:

• Publish AI agents on a marketplace for global discovery
• Receive compensation for the AI agents they create
• Rent agents on demand to accelerate the development of AI startups by 10x

Coral Server (Runtime)

Coral Server implements MCP primitives, allowing agents to register, create threads, send messages, and mention other agents. This enables structured agent-to-agent coordination, reducing the need for brittle context splicing.

Coral CLI + Studio

This tool allows developers to add remote/local agents, integrate them into shared threads, and inspect thread/message telemetry for debugging and performance optimization.

Registry Surface

The registry serves as a discovery layer for finding and integrating agents. Monetization and hosted checkout features are explicitly marked as «coming soon.»

Why Interoperability Matters

Current agent frameworks, such as LangChain and CrewAI, lack a common operational protocol, hindering composition. Coral’s MCP threading model provides a unified transport and addressing scheme, enabling specialized agents to coordinate without the need for ad-hoc glue code or prompt concatenation. The Coral Protocol team emphasizes the importance of persistent threads and mention-based targeting to maintain organized collaboration with minimal overhead.

Reference Implementation: Anemoi on GAIA

Coral’s open implementation, Anemoi, showcases a semi-centralized pattern: a light planner and specialized workers communicating directly via Coral MCP threads. On GAIA, Anemoi achieved a 52.73% pass@3 using GPT-4.1-mini (planner) and GPT-4o (workers), outperforming a reproduced OWL setup at 43.63% under identical LLM/tooling conditions. The arXiv paper and GitHub readme document these results and the coordination loop (plan → execute → critique → refine).

This design reduces reliance on a single powerful planner, minimizes redundant token passing, and enhances scalability and cost-effectiveness for long-horizon tasks. The evidence suggests that structured agent-to-agent communication is superior to naive prompt chaining when planner capacity is limited.

Incentives and Marketplace Status

Coral is positioning itself as a usage-based marketplace where agent authors can list agents with pricing metadata and receive payment per call. As of now, the developer page clearly indicates that «Pay Per Usage / Get Paid Automatically» and «Hosted checkout» features are coming soon. Teams should refrain from assuming general availability for payouts until Coral provides an update.

Summary

Coral v1 offers a standards-first interoperability runtime for multi-agent systems, along with practical tools for discovery and observability. The Anemoi GAIA results provide empirical support for the thread-based design under constrained planners. While the marketplace narrative is compelling, it is advisable to treat monetization as forthcoming, as indicated by Coral’s own communications. Developers should focus on building against the runtime and registry now, keeping payment features flagged until they are generally available.

Introducing Coral v1.

For the first time, anyone can:

• Publish AI agents on a marketplace where the world can discover them
• Get paid for AI agents they create
• Rent agents on demand to build AI startups 10x faster

Here’s why this matters.