I've been following Google's A2A protocol since it launched and noticed that there is still no good way to find agents out on the public internet. They're scattered across GitHub repos, registries, cloud deployments, random subdomains, and many go offline without anyone noticing.
So I built Waggle, which is a search engine that crawls the web for any domain that exposes a valid agent card, indexes them with semantic embeddings, and tracks their health over time. Waggle's index is exposed in a few different ways. First, there is good old search. There is also a REST API for programmatic access. But the feature I'm most excited about is Waggle's A2A-compliant meta-agent that knows how to delegate tasks to other agents. You ask Waggle "What are the GPS coordinates of the Empire State Building", it queries the database, finds a geocoding agent, and hands off your task, then gets back to you when the other agent is done.
The ecosystem is still very young and a lot of agents my crawlers dredge up are still "hello world" demos that are neither useful nor properly implemented. But I'm happy to report that over the course of one month, I went from 8 indexed agents to over 100, with about half being online at any given time. You can try out the Waggle agent / task delegation by using the chat feature (no registration required).
Some known-good agents you can take for a spin:\n- Cliff the Surveyor (geocoding, earthquake / flood analysis in the US)\n- Dispute_Email_Agent (functions as advertised)\n- OpSpawn AI Agent (Tries to do many things, I only had luck with converting markdown to HTML)
In any case, excited to share! Happy to answer questions."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Show HN: Waggle \u2013 A search engine for A2A protocol agents"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://waggle.zone"}},"_tags":["story","author_enmerk4r","story_47138064","show_hn"],"author":"enmerk4r","created_at":"2026-02-24T15:05:12Z","created_at_i":1771945512,"num_comments":0,"objectID":"47138064","points":1,"story_id":47138064,"story_text":"Hey HN,
I've been following Google's A2A protocol since it launched and noticed that there is still no good way to find agents out on the public internet. They're scattered across GitHub repos, registries, cloud deployments, random subdomains, and many go offline without anyone noticing.
So I built Waggle, which is a search engine that crawls the web for any domain that exposes a valid agent card, indexes them with semantic embeddings, and tracks their health over time. Waggle's index is exposed in a few different ways. First, there is good old search. There is also a REST API for programmatic access. But the feature I'm most excited about is Waggle's A2A-compliant meta-agent that knows how to delegate tasks to other agents. You ask Waggle "What are the GPS coordinates of the Empire State Building", it queries the database, finds a geocoding agent, and hands off your task, then gets back to you when the other agent is done.
The ecosystem is still very young and a lot of agents my crawlers dredge up are still "hello world" demos that are neither useful nor properly implemented. But I'm happy to report that over the course of one month, I went from 8 indexed agents to over 100, with about half being online at any given time. You can try out the Waggle agent / task delegation by using the chat feature (no registration required).
Some known-good agents you can take for a spin:\n- Cliff the Surveyor (geocoding, earthquake / flood analysis in the US)\n- Dispute_Email_Agent (functions as advertised)\n- OpSpawn AI Agent (Tries to do many things, I only had luck with converting markdown to HTML)
In any case, excited to share! Happy to answer questions.","title":"Show HN: Waggle \u2013 A search engine for A2A protocol agents","updated_at":"2026-03-05T23:36:25Z","url":"https://waggle.zone"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ferasawady"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"I've been building MXP, a high-performance protocol for AI agent communication.
Benchmarks (verified, reproducible):
256-byte message encode: 60ns (MXP) vs 2,262ns (JSON)\nBuilt-in trace IDs in every message (no OpenTelemetry instrumentation needed)\nNative streaming for LLM tokens\nWritten in Rust, MIT/Apache-2.0 licensed, public domain spec.
Looking for feedback on:
The "enabler not competitor" positioning\nWhat SDKs to prioritize (JS is next, then Python)\nWhether A2A compatibility matters to you\nRepo: https://github.com/yafatek/mxp-protocol"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Show HN: MXP \u2013 A2A-compatible agent protocol, 37x faster than JSON"}},"_tags":["story","author_ferasawady","story_46071309","show_hn"],"author":"ferasawady","created_at":"2025-11-27T17:22:06Z","created_at_i":1764264126,"num_comments":0,"objectID":"46071309","points":1,"story_id":46071309,"story_text":"I've been building MXP, a high-performance protocol for AI agent communication.
Benchmarks (verified, reproducible):
256-byte message encode: 60ns (MXP) vs 2,262ns (JSON)\nBuilt-in trace IDs in every message (no OpenTelemetry instrumentation needed)\nNative streaming for LLM tokens\nWritten in Rust, MIT/Apache-2.0 licensed, public domain spec.
Looking for feedback on:
The "enabler not competitor" positioning\nWhat SDKs to prioritize (JS is next, then Python)\nWhether A2A compatibility matters to you\nRepo: https://github.com/yafatek/mxp-protocol","title":"Show HN: MXP \u2013 A2A-compatible agent protocol, 37x faster than JSON","updated_at":"2026-03-05T23:09:09Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Zavora"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Hey everyone!
I'm excited to share ADK-Rust - a production-ready implementation of Google's Agent Development Kit in Rust.
Why Rust?\nAfter working extensively with adk-python in developing an ai agent factory at zavora.ai, I wanted to bring the same powerful agent development patterns to the Rust ecosystem, targeting use cases where:
Performance is critical - Rust's zero-cost abstractions and memory safety\nDeployment size matters - Single binary with no runtime dependencies\nSystems-level integration - Embedded systems, edge computing, IoT\nConcurrency at scale - Rust's async/await with tokio\nFeatures\nADK-Rust maintains API parity with the Python ADK where possible:
Model-agnostic - Gemini, OpenAI, Anthropic, DeepSeek support\n Multiple agent types - LlmAgent, SequentialAgent, ParallelAgent, LoopAgent\n Tool support - Built-in tools (Google Search, Code Execution) + custom tools\n MCP support - Model Context Protocol integration\n Sessions & Memory - InMemorySessionService, DatabaseSessionService\n Streaming - Full streaming support for real-time responses\n Telemetry - OpenTelemetry integration for tracing/metrics\n A2A Protocol - Agent-to-Agent communication
Quick Example
use adk_rust::prelude::*;
#[tokio::main]\nasync fn main() -> Result<()> {\n let agent = LlmAgentBuilder::new()\n .name("my_agent")\n .model(GeminiModel::new("gemini-2.0-flash")?)\n .instruction("You are a helpful assistant.")\n .build()?;
let response = agent.run("Hello!").await?;\n println!("{}", response);\n Ok(())Links\n Crates.io: https://crates.io/crates/adk-rust\n Docs: https://docs.rs/adk-rust\n Website: https://adk-rust.com/\n GitHub: https://github.com/zavora-ai/adk-rust\nLooking for Feedback\nI'd love to hear from the community:
What agentic features would you prioritize?\nAny interest in contributing or testing?\nUse cases where a Rust implementation would be valuable?\nThis is an independent community project, not officially affiliated with Google, but designed to be compatible with the ADK ecosystem.
Thanks for reading!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["agent"],"value":"Show HN: ADK-Rust: a Rust Implementation of Google Agent Dev Kit"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://adk-rust.com"}},"_tags":["story","author_Zavora","story_46252336","show_hn"],"author":"Zavora","created_at":"2025-12-13T05:45:09Z","created_at_i":1765604709,"num_comments":0,"objectID":"46252336","points":2,"story_id":46252336,"story_text":"Hey everyone!
I'm excited to share ADK-Rust - a production-ready implementation of Google's Agent Development Kit in Rust.
Why Rust?\nAfter working extensively with adk-python in developing an ai agent factory at zavora.ai, I wanted to bring the same powerful agent development patterns to the Rust ecosystem, targeting use cases where:
Performance is critical - Rust's zero-cost abstractions and memory safety\nDeployment size matters - Single binary with no runtime dependencies\nSystems-level integration - Embedded systems, edge computing, IoT\nConcurrency at scale - Rust's async/await with tokio\nFeatures\nADK-Rust maintains API parity with the Python ADK where possible:
Model-agnostic - Gemini, OpenAI, Anthropic, DeepSeek support\n Multiple agent types - LlmAgent, SequentialAgent, ParallelAgent, LoopAgent\n Tool support - Built-in tools (Google Search, Code Execution) + custom tools\n MCP support - Model Context Protocol integration\n Sessions & Memory - InMemorySessionService, DatabaseSessionService\n Streaming - Full streaming support for real-time responses\n Telemetry - OpenTelemetry integration for tracing/metrics\n A2A Protocol - Agent-to-Agent communication
Quick Example
use adk_rust::prelude::*;
#[tokio::main]\nasync fn main() -> Result<()> {\n let agent = LlmAgentBuilder::new()\n .name("my_agent")\n .model(GeminiModel::new("gemini-2.0-flash")?)\n .instruction("You are a helpful assistant.")\n .build()?;
let response = agent.run("Hello!").await?;\n println!("{}", response);\n Ok(())Links\n Crates.io: https://crates.io/crates/adk-rust\n Docs: https://docs.rs/adk-rust\n Website: https://adk-rust.com/\n GitHub: https://github.com/zavora-ai/adk-rust\nLooking for Feedback\nI'd love to hear from the community:
What agentic features would you prioritize?\nAny interest in contributing or testing?\nUse cases where a Rust implementation would be valuable?\nThis is an independent community project, not officially affiliated with Google, but designed to be compatible with the ADK ecosystem.
Thanks for reading!","title":"Show HN: ADK-Rust: a Rust Implementation of Google Agent Dev Kit","updated_at":"2026-03-05T23:10:57Z","url":"https://adk-rust.com"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kittikitti"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"A2A is a very high level protocol that's meant to be an entry point for the actual network protocols that allow agents to communicate with others like HTTP, gRPC, and others. The promise of no-code means the concept has to exist as an abstraction.
There's a Kaggle course going on where Google discusses A2A and their thoughts on it. If you're a pure vibe coder who doesn't know how to code, A2A is for you. If you know the basics of programming and could even do some web crawling, there are many accessible options.
Google and the industry is honing in on these vibe coders who will look at 10 million tokens consumed to make a checklist application and think nothing of it. The agent to agent (A2A) protocol is for them. Personally, I think it's useful to describe what I'm already doing to people who aren't experts."},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a","protocol"],"value":"Ask HN: Is anyone using the A2A protocol?"}},"_tags":["comment","author_kittikitti","story_48582679"],"author":"kittikitti","comment_text":"A2A is a very high level protocol that's meant to be an entry point for the actual network protocols that allow agents to communicate with others like HTTP, gRPC, and others. The promise of no-code means the concept has to exist as an abstraction.
There's a Kaggle course going on where Google discusses A2A and their thoughts on it. If you're a pure vibe coder who doesn't know how to code, A2A is for you. If you know the basics of programming and could even do some web crawling, there are many accessible options.
Google and the industry is honing in on these vibe coders who will look at 10 million tokens consumed to make a checklist application and think nothing of it. The agent to agent (A2A) protocol is for them. Personally, I think it's useful to describe what I'm already doing to people who aren't experts.","created_at":"2026-06-18T20:51:09Z","created_at_i":1781815869,"objectID":"48591379","parent_id":48590772,"story_id":48582679,"story_title":"Ask HN: Is anyone using the A2A protocol?","updated_at":"2026-06-19T04:53:43Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"el1fe"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Hey HN! I built Cogitator\u2014 an open-source, self-hosted AI agent runtime for TypeScript. Think of it as "Kubernetes for AI Agents."
We just shipped native support for Google's A2A Protocol (Agent-to-Agent) v0.3 \u2014 implemented from scratch, zero external dependencies, ~1500 lines of code with 119 tests. As far as I know, this is the first TypeScript runtime with a native A2A implementation.
What A2A enables: Your agents can talk to agents from other frameworks (LangChain, CrewAI, etc.) via a standardized JSON-RPC protocol. You expose an agent as an A2A server, and anyone can discover it via Agent Cards and send it tasks \u2014 with streaming support.
The killer feature is asTool() \u2014 wrap any remote A2A agent as a local tool:
const client = new A2AClient('https://research-agent.example.com');\n const remoteTool = client.asToolFromCard(await client.agentCard());\n const agent = new Agent({ tools: [remoteTool] });\nEverything is TypeScript-native, self-hosted, no vendor lock-in. MIT licensed.
npm (core): https://www.npmjs.com/package/@cogitator-ai/core\nnpm (a2a): https://www.npmjs.com/package/@cogitator-ai/a2a
Happy to answer any questions about the architecture or A2A implementation!"},"story_title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Show HN: Cogitator \u2013 Self-hosted AI agent runtime with native A2A Protocol"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/cogitator-ai/Cogitator-AI"}},"_tags":["comment","author_el1fe","story_47078206"],"author":"el1fe","comment_text":"Hey HN! I built Cogitator\u2014 an open-source, self-hosted AI agent runtime for TypeScript. Think of it as "Kubernetes for AI Agents."
We just shipped native support for Google's A2A Protocol (Agent-to-Agent) v0.3 \u2014 implemented from scratch, zero external dependencies, ~1500 lines of code with 119 tests. As far as I know, this is the first TypeScript runtime with a native A2A implementation.
What A2A enables: Your agents can talk to agents from other frameworks (LangChain, CrewAI, etc.) via a standardized JSON-RPC protocol. You expose an agent as an A2A server, and anyone can discover it via Agent Cards and send it tasks \u2014 with streaming support.
The killer feature is asTool() \u2014 wrap any remote A2A agent as a local tool:
const client = new A2AClient('https://research-agent.example.com');\n const remoteTool = client.asToolFromCard(await client.agentCard());\n const agent = new Agent({ tools: [remoteTool] });\nEverything is TypeScript-native, self-hosted, no vendor lock-in. MIT licensed.
npm (core): https://www.npmjs.com/package/@cogitator-ai/core\nnpm (a2a): https://www.npmjs.com/package/@cogitator-ai/a2a
Happy to answer any questions about the architecture or A2A implementation!","created_at":"2026-02-19T19:47:38Z","created_at_i":1771530458,"objectID":"47078228","parent_id":47078206,"story_id":47078206,"story_title":"Show HN: Cogitator \u2013 Self-hosted AI agent runtime with native A2A Protocol","story_url":"https://github.com/cogitator-ai/Cogitator-AI","updated_at":"2026-03-05T23:35:57Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sanket-msd"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Cisco AI Defense, open-sourced the A2A Scanner, a tool designed to secure Agent-to-Agent (A2A) protocol ecosystem. It helps validate agent identity, detect spoofing and prompt injection risks, and includes modular analyzers that anyone can extend or customize. If you are exploring agentic AI frameworks or building multi-agent systems, we would love feedback, contributions, and critique from the community. Repo link: https://github.com/cisco-ai-defense/a2a-scanner.\nBlog link: https://blogs.cisco.com/ai/securing-ai-agents-with-ciscos-op..."},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a"],"value":"Cisco AI's A2A Scanner \u2013 New OSS Tool to Detect Threats in A2A Eco-System"},"story_url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a"],"value":"https://github.com/cisco-ai-defense/a2a-scanner"}},"_tags":["comment","author_sanket-msd","story_46209497"],"author":"sanket-msd","comment_text":"Cisco AI Defense, open-sourced the A2A Scanner, a tool designed to secure Agent-to-Agent (A2A) protocol ecosystem. It helps validate agent identity, detect spoofing and prompt injection risks, and includes modular analyzers that anyone can extend or customize. If you are exploring agentic AI frameworks or building multi-agent systems, we would love feedback, contributions, and critique from the community. Repo link: https://github.com/cisco-ai-defense/a2a-scanner.\nBlog link: https://blogs.cisco.com/ai/securing-ai-agents-with-ciscos-op...","created_at":"2025-12-09T19:36:14Z","created_at_i":1765308974,"objectID":"46209498","parent_id":46209497,"story_id":46209497,"story_title":"Cisco AI's A2A Scanner \u2013 New OSS Tool to Detect Threats in A2A Eco-System","story_url":"https://github.com/cisco-ai-defense/a2a-scanner","updated_at":"2026-03-05T23:07:38Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sascha_heyer"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"I've been following the OpenAI Agents repo and came across this pull request.
A contributor submitted a massive, 1,200+ line PR to add support for the Agent-to-Agent (A2A) protocol. This wasn't a lazy contribution.
It included 30 tests, full documentation, and core infrastructure for agent interoperability.
What's notable is that the A2A protocol was initiated and specified by Google.
The response from the OpenAI maintainer showed zero appreciation for the significant effort. It was a cold, procedural shutdown:
"Hi @Kunmeer-SyedMohamedHyder, thanks for sending this pull request. However, we don't have immediate plans to add A2A support to this SDK. We can't say if or when we'll review it. For future major contributions, it would be appreciated if you could start with an issue for discussion first."
The PR was then marked as draft and auto-closed by a bot for inactivity.
This is a perfect example of the "open source" facade from large AI labs. An engineer does a huge amount of high-quality, free work to add a valuable, open interoperability standard. Instead of appreciation or discussion, they're met with a "we don't have plans" and a passive-aggressive "ask first next time."
It's hard not to see this as a "Not Invented By A Competitor" rejection. Why else would you show zero interest in a feature that so clearly benefits the entire agent ecosystem? It sends a strong message to any potential contributor: don't bother building anything that doesn't align with OpenAI's closed roadmap, especially if it has Google's name on it.
What's the point of hosting on GitHub if you treat high-effort community contributions like this?"},"story_title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"OpenAI rejects 1,200-line community PR for Google's A2A agent protocol"},"story_url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["agent"],"value":"https://github.com/openai/openai-agents-python/pull/1245"}},"_tags":["comment","author_sascha_heyer","story_45766384"],"author":"sascha_heyer","comment_text":"I've been following the OpenAI Agents repo and came across this pull request.
A contributor submitted a massive, 1,200+ line PR to add support for the Agent-to-Agent (A2A) protocol. This wasn't a lazy contribution.
It included 30 tests, full documentation, and core infrastructure for agent interoperability.
What's notable is that the A2A protocol was initiated and specified by Google.
The response from the OpenAI maintainer showed zero appreciation for the significant effort. It was a cold, procedural shutdown:
"Hi @Kunmeer-SyedMohamedHyder, thanks for sending this pull request. However, we don't have immediate plans to add A2A support to this SDK. We can't say if or when we'll review it. For future major contributions, it would be appreciated if you could start with an issue for discussion first."
The PR was then marked as draft and auto-closed by a bot for inactivity.
This is a perfect example of the "open source" facade from large AI labs. An engineer does a huge amount of high-quality, free work to add a valuable, open interoperability standard. Instead of appreciation or discussion, they're met with a "we don't have plans" and a passive-aggressive "ask first next time."
It's hard not to see this as a "Not Invented By A Competitor" rejection. Why else would you show zero interest in a feature that so clearly benefits the entire agent ecosystem? It sends a strong message to any potential contributor: don't bother building anything that doesn't align with OpenAI's closed roadmap, especially if it has Google's name on it.
What's the point of hosting on GitHub if you treat high-effort community contributions like this?","created_at":"2025-10-30T22:57:30Z","created_at_i":1761865050,"objectID":"45766385","parent_id":45766384,"story_id":45766384,"story_title":"OpenAI rejects 1,200-line community PR for Google's A2A agent protocol","story_url":"https://github.com/openai/openai-agents-python/pull/1245","updated_at":"2026-03-05T22:58:55Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ismats"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Google's A2A Protocol: How AI Agents Will Talk to Each Other"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"https://www.ismatsamadov.com/blog/a2a-protocol-agent-to-agent-google"}},"_tags":["story","author_ismats","story_47923422"],"author":"ismats","created_at":"2026-04-27T16:07:58Z","created_at_i":1777306078,"num_comments":0,"objectID":"47923422","points":2,"story_id":47923422,"title":"Google's A2A Protocol: How AI Agents Will Talk to Each Other","updated_at":"2026-04-27T16:22:27Z","url":"https://www.ismatsamadov.com/blog/a2a-protocol-agent-to-agent-google"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kadirpekel"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Hey HN,
I've been building Hector, a declarative AI agent platform in Go that uses the A2A protocol. The idea is pretty simple: instead of writing code to build agents, you just define everything in YAML.
Want to create an agent? Write a YAML file with the prompt, reasoning strategy, tools, and you're done. No Python, no SDKs, no complex setup. It's like infrastructure as code but for AI agents.
The cool part is that since it's built on A2A (Agent-to-Agent protocol), agents can talk to each other seamlessly. You can mix local agents with remote ones, or have agents from different systems work together. It's kind of like Docker for AI agents.
I built this because I got tired of the complexity in current agent frameworks. Most require you to write a bunch of boilerplate code just to get started. With Hector, you focus on the logic, not the plumbing.
It's still in alpha, but the core stuff works. I'd love to get feedback from anyone working on agentic systems or multi-agent coordination. What pain points do you see in current approaches?
Repo: https://github.com/kadirpekel/hector
Would appreciate any thoughts or feedback!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a","agent"],"value":"Show HN: Hector \u2013 Pure A2A-Native Declarative AI Agent Platform (Go)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/kadirpekel/hector"}},"_tags":["story","author_kadirpekel","story_45493411","show_hn"],"author":"kadirpekel","created_at":"2025-10-06T16:58:46Z","created_at_i":1759769926,"num_comments":0,"objectID":"45493411","points":3,"story_id":45493411,"story_text":"Hey HN,
I've been building Hector, a declarative AI agent platform in Go that uses the A2A protocol. The idea is pretty simple: instead of writing code to build agents, you just define everything in YAML.
Want to create an agent? Write a YAML file with the prompt, reasoning strategy, tools, and you're done. No Python, no SDKs, no complex setup. It's like infrastructure as code but for AI agents.
The cool part is that since it's built on A2A (Agent-to-Agent protocol), agents can talk to each other seamlessly. You can mix local agents with remote ones, or have agents from different systems work together. It's kind of like Docker for AI agents.
I built this because I got tired of the complexity in current agent frameworks. Most require you to write a bunch of boilerplate code just to get started. With Hector, you focus on the logic, not the plumbing.
It's still in alpha, but the core stuff works. I'd love to get feedback from anyone working on agentic systems or multi-agent coordination. What pain points do you see in current approaches?
Repo: https://github.com/kadirpekel/hector
Would appreciate any thoughts or feedback!","title":"Show HN: Hector \u2013 Pure A2A-Native Declarative AI Agent Platform (Go)","updated_at":"2026-03-05T22:51:26Z","url":"https://github.com/kadirpekel/hector"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"oscarsixsecllc"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Hey HN,
I'm Randy, founder of Oscar Six Security. We build vulnerability scanning tools for small businesses and MSPs.
Today I'm sharing something I'm genuinely excited about: we shipped native support for Google's A2A (Agent-to-Agent) protocol in our Radar scanner. Here's what that actually means in practice:
An AI agent \u2014 anything built on Google ADK, LangChain, CrewAI, or any A2A-compatible framework \u2014 can now:
1. Discover Radar's capabilities via our agent card at /.well-known/agent.json\n2. Initiate payment via Stripe SPT or saved payment method (no human credit card entry)\n3. Submit a domain for scanning\n4. Poll for results and receive a structured vulnerability report
The entire vulnerability scanning lifecycle, with no human in the loop.
We handle domain verification before any scan runs \u2014 DNS challenge or web-based fallback \u2014 which an agent can resolve programmatically. Tier 1 pre-verified domains skip verification entirely.
Why build this now? We think security tooling is about to go agent-native fast. If your SOC automation or compliance pipeline runs on AI agents, having to context-switch to a browser portal to run a scan is a UX cliff. We're removing that cliff.
Scans are $99. The A2A endpoint is live today.
Technical details: JSON-RPC 2.0, agent card at /.well-known/agent.json, tiered domain verification (pre-verified / DNS challenge / web fallback).
Blog post with full technical walkthrough:\nhttps://blog.oscarsixsecurityllc.com/blog/oscar-six-radar-a2a-agent-to-agent-vulnerability-scanning?utm_source=hackernews&utm_medium=social&utm_campaign=a2a_announcement
Platform:\nhttps://radar.oscarsixsecurityllc.com/?utm_source=hackernews&utm_medium=social&utm_campaign=a2a_announcement
Happy to go deep on the implementation, the A2A spec, or the domain verification design. Ask anything."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a","agent"],"value":"Oscar Six Radar \u2013 vulnerability scanner with native A2A (agent-to-agent) support"}},"_tags":["story","author_oscarsixsecllc","story_47242324","ask_hn"],"author":"oscarsixsecllc","created_at":"2026-03-04T02:40:41Z","created_at_i":1772592041,"num_comments":0,"objectID":"47242324","points":1,"story_id":47242324,"story_text":"Hey HN,
I'm Randy, founder of Oscar Six Security. We build vulnerability scanning tools for small businesses and MSPs.
Today I'm sharing something I'm genuinely excited about: we shipped native support for Google's A2A (Agent-to-Agent) protocol in our Radar scanner. Here's what that actually means in practice:
An AI agent \u2014 anything built on Google ADK, LangChain, CrewAI, or any A2A-compatible framework \u2014 can now:
1. Discover Radar's capabilities via our agent card at /.well-known/agent.json\n2. Initiate payment via Stripe SPT or saved payment method (no human credit card entry)\n3. Submit a domain for scanning\n4. Poll for results and receive a structured vulnerability report
The entire vulnerability scanning lifecycle, with no human in the loop.
We handle domain verification before any scan runs \u2014 DNS challenge or web-based fallback \u2014 which an agent can resolve programmatically. Tier 1 pre-verified domains skip verification entirely.
Why build this now? We think security tooling is about to go agent-native fast. If your SOC automation or compliance pipeline runs on AI agents, having to context-switch to a browser portal to run a scan is a UX cliff. We're removing that cliff.
Scans are $99. The A2A endpoint is live today.
Technical details: JSON-RPC 2.0, agent card at /.well-known/agent.json, tiered domain verification (pre-verified / DNS challenge / web fallback).
Blog post with full technical walkthrough:\nhttps://blog.oscarsixsecurityllc.com/blog/oscar-six-radar-a2a-agent-to-agent-vulnerability-scanning?utm_source=hackernews&utm_medium=social&utm_campaign=a2a_announcement
Platform:\nhttps://radar.oscarsixsecurityllc.com/?utm_source=hackernews&utm_medium=social&utm_campaign=a2a_announcement
Happy to go deep on the implementation, the A2A spec, or the domain verification design. Ask anything.","title":"Oscar Six Radar \u2013 vulnerability scanner with native A2A (agent-to-agent) support","updated_at":"2026-03-05T23:41:07Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"xmas123"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"I built Entangl, an open-source post-quantum A2A (agent-to-agent) communication protocol for AI agents.
The problem: agents are negotiating contracts, executing transactions, and passing sensitive payloads between themselves. All of that traffic is protected by RSA and ECDH \u2014 algorithms that Shor's algorithm breaks on a sufficiently large quantum computer. Nation-state adversaries run harvest-now-decrypt-later attacks today.
What Entangl does:\n- CRYSTALS-Kyber1024 (NIST FIPS 203) replaces RSA/ECDH key exchange\n- CRYSTALS-Dilithium5 (NIST FIPS 204) replaces ECDSA signatures\n- Fresh KEM per message \u2014 forward secrecy, no session keys to steal\n- Each agent has a DID tethered to a human owner for accountability\n- Optional BB84 QKD layer via Cirq \u2014 eavesdroppers detectable at ~25% QBER\n- Routing server forwards encrypted envelopes but cannot read them
Demo: two agents negotiate a GPU compute deal in 1.4s over a live WebSocket server. Rogue agents blocked at registry level. Tampered ciphertext caught by signature check.
Stack: Python 3.11, Cirq 1.3, TensorFlow Quantum 0.7.2, FastAPI, NVIDIA RTX A1000.
Feedback welcome \u2014 especially on the per-message KEM vs. session key with periodic rotation tradeoff."},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["protocol"],"value":"Entangl \u2013 Post-quantum secure communication protocol for AI agents"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/amitb-quantum/entangl"}},"_tags":["comment","author_xmas123","story_47351983"],"author":"xmas123","children":[47354317],"comment_text":"I built Entangl, an open-source post-quantum A2A (agent-to-agent) communication protocol for AI agents.
The problem: agents are negotiating contracts, executing transactions, and passing sensitive payloads between themselves. All of that traffic is protected by RSA and ECDH \u2014 algorithms that Shor's algorithm breaks on a sufficiently large quantum computer. Nation-state adversaries run harvest-now-decrypt-later attacks today.
What Entangl does:\n- CRYSTALS-Kyber1024 (NIST FIPS 203) replaces RSA/ECDH key exchange\n- CRYSTALS-Dilithium5 (NIST FIPS 204) replaces ECDSA signatures\n- Fresh KEM per message \u2014 forward secrecy, no session keys to steal\n- Each agent has a DID tethered to a human owner for accountability\n- Optional BB84 QKD layer via Cirq \u2014 eavesdroppers detectable at ~25% QBER\n- Routing server forwards encrypted envelopes but cannot read them
Demo: two agents negotiate a GPU compute deal in 1.4s over a live WebSocket server. Rogue agents blocked at registry level. Tampered ciphertext caught by signature check.
Stack: Python 3.11, Cirq 1.3, TensorFlow Quantum 0.7.2, FastAPI, NVIDIA RTX A1000.
Feedback welcome \u2014 especially on the per-message KEM vs. session key with periodic rotation tradeoff.","created_at":"2026-03-12T15:19:22Z","created_at_i":1773328762,"objectID":"47351984","parent_id":47351983,"story_id":47351983,"story_title":"Entangl \u2013 Post-quantum secure communication protocol for AI agents","story_url":"https://github.com/amitb-quantum/entangl","updated_at":"2026-03-12T17:29:23Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"MattDaEskimo"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"I'm concerned that this fits in "using today's innovation to solve outdated paradigms".
Google has A2A: An Agent-to-Agent Protocol. SaaS is plumetting in value.
Arbitrary semantics made sense when communications were human-dominated.
If agents dominate these fields, why wouldn't they simply set their own protocols and methods to communicate both text, binary, and agreed data structures?
There's an assumption that email is somehow the best channel, when you've found yourself that the most popular, functional interfaces don't align with your expectations.
Then, ultimately I have a single agent that can sit in numerous communication platforms, such as email"},"story_title":{"matchLevel":"none","matchedWords":[],"value":"Launch HN: AgentMail (YC S25) \u2013 An API that gives agents their own email inboxes"}},"_tags":["comment","author_MattDaEskimo","story_46812608"],"author":"MattDaEskimo","children":[46814576,46830251],"comment_text":"I'm concerned that this fits in "using today's innovation to solve outdated paradigms".
Google has A2A: An Agent-to-Agent Protocol. SaaS is plumetting in value.
Arbitrary semantics made sense when communications were human-dominated.
If agents dominate these fields, why wouldn't they simply set their own protocols and methods to communicate both text, binary, and agreed data structures?
There's an assumption that email is somehow the best channel, when you've found yourself that the most popular, functional interfaces don't align with your expectations.
Then, ultimately I have a single agent that can sit in numerous communication platforms, such as email","created_at":"2026-01-29T18:30:41Z","created_at_i":1769711441,"objectID":"46814325","parent_id":46812608,"story_id":46812608,"story_title":"Launch HN: AgentMail (YC S25) \u2013 An API that gives agents their own email inboxes","updated_at":"2026-03-05T23:26:22Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"franzvill"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"AI agents are getting really good at doing things, but they're completely blind to their physical surroundings.
If you walk into a hotel and you have an AI assistant (like the Chatgpt mobile app), it has no idea there may be a concierge agent on the network that could help you book a spa, check breakfast times, or request late checkout. Same thing at offices, hospitals, cruise ships. The agents are there, but there's no way to discover them.
A2A (Google's agent-to-agent protocol) handles how agents talk to each other. MCP handles how agents use tools. But neither answers a basic question: how do you find agents in the first place?
So I built LAD-A2A, a simple discovery protocol. When you connect to a Wi-Fi, your agent can automatically find what's available using mDNS (like how AirDrop finds nearby devices) or a standard HTTP endpoint.
The spec is intentionally minimal. I didn't want to reinvent A2A or create another complex standard. LAD-A2A just handles discovery, then hands off to A2A for actual communication.
Open source, Apache 2.0. Includes a working Python implementation you can run to see it in action.
Curious what people think!
https://lad-a2a.org/\nhttps://github.com/franzvill/lad"},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a","agent"],"value":"Lad-A2A: How AI agents find each other on local networks"},"story_url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a"],"value":"https://lad-a2a.org/"}},"_tags":["comment","author_franzvill","story_46801380"],"author":"franzvill","children":[46801496],"comment_text":"AI agents are getting really good at doing things, but they're completely blind to their physical surroundings.
If you walk into a hotel and you have an AI assistant (like the Chatgpt mobile app), it has no idea there may be a concierge agent on the network that could help you book a spa, check breakfast times, or request late checkout. Same thing at offices, hospitals, cruise ships. The agents are there, but there's no way to discover them.
A2A (Google's agent-to-agent protocol) handles how agents talk to each other. MCP handles how agents use tools. But neither answers a basic question: how do you find agents in the first place?
So I built LAD-A2A, a simple discovery protocol. When you connect to a Wi-Fi, your agent can automatically find what's available using mDNS (like how AirDrop finds nearby devices) or a standard HTTP endpoint.
The spec is intentionally minimal. I didn't want to reinvent A2A or create another complex standard. LAD-A2A just handles discovery, then hands off to A2A for actual communication.
Open source, Apache 2.0. Includes a working Python implementation you can run to see it in action.
Curious what people think!
https://lad-a2a.org/\nhttps://github.com/franzvill/lad","created_at":"2026-01-28T20:56:53Z","created_at_i":1769633813,"objectID":"46801381","parent_id":46801380,"story_id":46801380,"story_title":"Lad-A2A: How AI agents find each other on local networks","story_url":"https://lad-a2a.org/","updated_at":"2026-03-05T23:25:06Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"vishyouluck"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"This project is a google A2A (agent to agent) client implemented using WebSocket and Java. It facilitates real-time communication between agents by leveraging WebSocket. Also provides mechanism for agentic mesh and agentic rag and create complete worklow using a2a or mcp servers
A2AJava is a pure Java implementation designed to support both the Google A2A (Agent-to-Agent) protocol and the MCP (Model Context Protocol), making it a versatile framework for building agentic applications. The A2A protocol standardizes how AI agents communicate and collaborate across different systems, while the MCP protocol focuses on enabling agents to access and trigger a wide range of tools and resources."},"story_title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Online A2A Client for Google Agent to Agent Protocol"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://vishalmysore-a2aclient.hf.space/"}},"_tags":["comment","author_vishyouluck","story_44323060"],"author":"vishyouluck","comment_text":"This project is a google A2A (agent to agent) client implemented using WebSocket and Java. It facilitates real-time communication between agents by leveraging WebSocket. Also provides mechanism for agentic mesh and agentic rag and create complete worklow using a2a or mcp servers
A2AJava is a pure Java implementation designed to support both the Google A2A (Agent-to-Agent) protocol and the MCP (Model Context Protocol), making it a versatile framework for building agentic applications. The A2A protocol standardizes how AI agents communicate and collaborate across different systems, while the MCP protocol focuses on enabling agents to access and trigger a wide range of tools and resources.","created_at":"2025-06-19T22:20:28Z","created_at_i":1750371628,"objectID":"44323061","parent_id":44323060,"story_id":44323060,"story_title":"Online A2A Client for Google Agent to Agent Protocol","story_url":"https://vishalmysore-a2aclient.hf.space/","updated_at":"2025-06-19T23:42:03Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"vishyouluck"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"This project demonstrates the integration of Google's A2A (Agent to Agent) protocol with Selenium for automated web browsing, test validation, screen capture, and other web-related tasks. The implementation provides flexible deployment options and powerful multi-agent workflow capabilities."},"story_title":{"matchLevel":"none","matchedWords":[],"value":"[dead]"}},"_tags":["comment","author_vishyouluck","story_43955857"],"author":"vishyouluck","comment_text":"This project demonstrates the integration of Google's A2A (Agent to Agent) protocol with Selenium for automated web browsing, test validation, screen capture, and other web-related tasks. The implementation provides flexible deployment options and powerful multi-agent workflow capabilities.","created_at":"2025-05-11T18:31:10Z","created_at_i":1746988270,"objectID":"43955858","parent_id":43955857,"story_id":43955857,"story_title":"[dead]","updated_at":"2025-05-11T18:36:12Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mneil"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"MCP (Model Context Protocol) is essentially a protocol for tool access. It defines a standard way for large language models to interact with various tools, data, and resources. Simply put, MCP enables AI to use various functionalities, much like how programmers call functions.
A2A (Agent-to-Agent Protocol) focuses on agent collaboration. It establishes ways for intelligent agents to discover, communicate, and cooperate with each other, allowing different AI systems to work together like human teams.
Taken straight from this article:
https://a2aprotocol.ai/blog/a2a-vs-mcp"},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["a2a"],"value":"Ask HN: Can someone ELI5 how Google's A2A is different from MCP?"}},"_tags":["comment","author_mneil","story_43676444"],"author":"mneil","children":[43681800],"comment_text":"MCP (Model Context Protocol) is essentially a protocol for tool access. It defines a standard way for large language models to interact with various tools, data, and resources. Simply put, MCP enables AI to use various functionalities, much like how programmers call functions.
A2A (Agent-to-Agent Protocol) focuses on agent collaboration. It establishes ways for intelligent agents to discover, communicate, and cooperate with each other, allowing different AI systems to work together like human teams.
Taken straight from this article:
https://a2aprotocol.ai/blog/a2a-vs-mcp","created_at":"2025-04-14T01:04:39Z","created_at_i":1744592679,"objectID":"43677131","parent_id":43676444,"story_id":43676444,"story_title":"Ask HN: Can someone ELI5 how Google's A2A is different from MCP?","updated_at":"2025-04-14T14:38:09Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"yankouskia"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Auto Agent Protocol \u2013 an A2A vertical for AI agents buying cars"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://autoagentprotocol.org/"}},"_tags":["story","author_yankouskia","story_48204856"],"author":"yankouskia","created_at":"2026-05-20T08:51:07Z","created_at_i":1779267067,"num_comments":0,"objectID":"48204856","points":2,"story_id":48204856,"title":"Auto Agent Protocol \u2013 an A2A vertical for AI agents buying cars","updated_at":"2026-05-20T09:07:31Z","url":"https://autoagentprotocol.org/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"yankouskia"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["a2a","agent","protocol"],"value":"Auto Agent Protocol \u2013 open A2A profile lets AI agents talk to car dealerships"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["agent","protocol"],"value":"https://github.com/auto-agent-protocol/auto-agent-protocol"}},"_tags":["story","author_yankouskia","story_47967405"],"author":"yankouskia","children":[47967406,47967797],"created_at":"2026-04-30T19:52:51Z","created_at_i":1777578771,"num_comments":1,"objectID":"47967405","points":1,"story_id":47967405,"title":"Auto Agent Protocol \u2013 open A2A profile lets AI agents talk to car dealerships","updated_at":"2026-05-01T11:58:55Z","url":"https://github.com/auto-agent-protocol/auto-agent-protocol"}],"hitsPerPage":20,"nbHits":185,"nbPages":10,"page":0,"params":"query=A2A+agent+protocol&advancedSyntax=true&analyticsTags=backend","processingTimeMS":28,"processingTimingsMS":{"_request":{"roundTrip":15},"afterFetch":{"format":{"highlighting":1,"total":1},"merge":{"mergeLoop":{"prepareNextHit":4,"total":4},"total":4},"total":4},"fetch":{"query":10,"scanning":11,"total":22},"total":28},"query":"A2A agent protocol","serverTimeMS":29}