The Internet of Agents Hackathon proved what's possible with Coral v1. 3,000+ devs built 130+ apps using reusable AI agents on our Solana-powered infra & won $100k in prizes. Big thanks to @solana @MistralAI @lovable_dev @crossmint @aimlapi @elevenlabsio @nebiusai. Recap 🧵 ⬇️

From Outrage, AI Songs, and EU Compliance: My Analysis of the Rising Demand for Transparent AI Systems Day 4 of Agent Trust 🔒, and today I’m looking into transparency, something that keeps coming up across governments, users, and developers. So let's look at this from the

We are heading to Neurips 2025 To present our paper: Anemoi: A Semi-Centralized Multi-agent System Based on Agent-to-Agent Communication MCP server from @Coral_Protocol Would love to meet anyone else there working on agents

With vertical scaling, you're limited to the capabilities and innovations made by standalone providers. Horizontal scaling unlocks a level of composability that agentic systems need to be practically useful in real-world business settings. Different models, frameworks, and

In most distributed agent systems, revoking access means hoping the message reaches every cache, peer, and registry before the token expires. Credentials linger. Capability declarations keep circulating. "Terminated" becomes probabilistic. Our architecture eliminates this

Which of the recent model releases have you been getting the most day-to-day productivity gains from? Real-world work settings almost always provide better insights than raw benchmarks alone. We'd love to hear your thoughts!

Your user asks about a specific property in their budget. Agent A: "Sorry, this property is no longer available. Let me know if there's anything else I can help with." Agent B: "That property was recently sold, but here's one 500m away in the same neighbourhood within your

This is the most impressed I’ve been with a new AI tool since Sora. Google Banana Pro is so good. Its editing abilities, when given to agents, unlock so many use cases. We have one graphic designer/editor who is always swamped with work, but now all I had to do was work with

Thanks to everyone who took part

Watch how Coral saves time and engineering resources while ensuring that the security and data protection requirements for high-stakes enterprise use cases are met.

We are live now

Context tradeoffs are a core tension of agent development that we think a lot about. Agents need context to complete tasks, but every additional data point you pass expands the attack surface. Give too little, the agent fails. Give too much, and you're propagating sensitive

Research from Ariel University found that up to 60% of multi-agent exchanges involve unintended data propagation. Coral's session isolation mitigates this risk entirely. In cases where agents need to share and retain sensitive information, attestations make it easy for

The paper proves that a secure architectural foundation is the path toward mission-critical autonomous coordination. 📄 Full research paper: https://t.co/Q4Y8pEpFiD 🔗 Start building with Coral:

Security is not a marketing angle. It's a first-principles design constraint. Agents will eventually negotiate, transact, audit, govern, collaborate, and enforce without humans in the loop. We are building for that world, not just today’s demos.

“CORAL effectively eliminates privilege persistence vulnerabilities for on-chain operations. Its immutable Solana ledger anchors privileges to tamper-proof smart contracts, providing instant verification and preventing both persistence and drift.” Identity + payments + access

Empirical evaluations found that “session isolation was perfect across all trials,” indicating that parallel agent tasks do not leak context. This is essential for: → workflow automation → enterprise data environments → autonomous multi-agent chains Economic model &

The role of verifiable execution & identity: A2A and ACP rely on network-trust + configuration correctness of optional security features. Whereas Coral introduces cryptographic identity + on-chain verification as baseline components. This moves agent execution closer to

The paper notes that most of the current ecosystem remains largely insecure, not due to lack of capability but due to missing protocol enforcement layers. This is the difference between: → “agents that can act” vs. → “agents businesses can rely on” In Summary: “CORAL stands

The paper analysed three emerging approaches for agent-to-agent communication: → A2A – by Google → ACP – Agent Communication Protocol → CORAL – Coral Protocol The focus wasn’t on UX or capabilities, but security posture, architecture, and protocol-level guarantees.

Why this research matters? Most agent demos today look impressive, but security expectations are low because they rarely operate in real business or multi-party environments. Once agents handle money, identity, decisions, or autonomous actions, security stops being optional.