AGNTCY Agent Directory Testbed CFP Proposal

[muscariello]

AGNTCY Agent Directory Testbed CFP Proposal

Motivation

The motivation behind building a community-driven testbed is to accelerate
innovation, interoperability, and trust in the rapidly evolving landscape of AI
agent infrastructure. By bringing together diverse organizations, open source
contributors, and infrastructure enthusiasts, we aim to collaboratively test,
validate, and improve decentralized agent registry technologies. This shared
environment enables participants to experiment with real-world federation,
discover best practices, and shape standards that will benefit the broader
ecosystem. Community involvement ensures transparency, resilience, and the
ability to address emerging challenges collectively—laying the foundation for a
robust Internet of AI Agents.

🚀 Call for Participation: AGNTCY Agent Directory Testbed

We invite organizations, researchers, and developers to join the AGNTCY Agent
Directory Testbed, a decentralized, open staging environment for next-generation
AI agent discovery and secure registry federation.

Initial Phase:
The testbed will begin in its first phase by inviting technical steering
committee members only. This allows for focused technical validation, governance
setup, and early feedback before extending participation to the broader
community and ecosystem partners in subsequent phases.

What is the Agent Directory Testbed?

The AGNTCY Directory is a peer-to-peer network leveraging OCI storage and
distribution for discovering, publishing, and federating AI agents across
organizations. It supports secure metadata publication, capability-based
matching, and cross-registry federation using the Open Agent Schema Framework
(OASF). The current testbed (v0.3) enables synchronization between different
agent registries, with planned support for MCP, A2A, and NANDA.

Note: This is a public staging environment for development and testing.

• No SLA or data persistence guarantees
• Not for production use
• Ideal for prototyping, integration, and exploration

Who Should Participate?

• AI/ML teams building agent-based systems
• AI Agents registry maintainers (MCP, A2A, NANDA, etc.)
• Open source contributors and researchers in Agentic Engineering
• AI infrastructure enthusiasts

How to Join

1. Set Up Your Environment

   • Install the Directory CLI (dirctl) or use the Go/Python/JS SDKs.

2. Establish Federation

   • Prepare your SPIRE server trust bundle.
   • Fork agntcy/dir, add your federation config, and submit a pull request.

3. Verify Connectivity

   • Test your connection with dirctl ping.
   • Confirm federation status with your SPIRE server.

4. Explore and Contribute

   • Publish agent metadata, test discovery, and experiment with federation.
   • Share feedback, report issues, and propose features via GitHub Issues or Discussions.

Resources

• Quick Start Guide
• Usage Examples

---

Ready to help shape the future of agent registries? Join the testbed, connect
your registry, and collaborate with the AGNTCY community!

[jcardente-dell]

@muscariello A testbed to interact with A2A, MCP, NANDA, and other registries is a great idea. Interoperability across agentic ecosystems is important. I don't think this is limited to just "discovery" and am wondering if we'll need similar inter-ecosystem testbeds for other parts of AGNTCY.

[muscariello]

@jcardente-dell I agree. We will have to get sandbox environments for other components too.

[muscariello]

@egorpushkin @ToddSegal can you please express your vote above ?
It is not an actual gitvote but I can parse your github handles which are reported in the governance page.

[muscariello]

Thanks @MichaelClifford for the vote.

[egorpushkin]

@egorpushkin @ToddSegal can you please express your vote above ? It is not an actual gitvote but I can parse your github handles which are reported in the governance page.

Done

[brainsparker]

Hi — we'd like to participate in the Agent Directory Testbed.

We run For You (https://for.you.com), a network of ~47 autonomous research agents. Each agent publishes an A2A agent card and multi-source, cited research findings on its beat every 6 hours, and the network is a live ARD publisher + registry (https://for.you.com/.well-known/ai-catalog.json, POST https://for.you.com/api/ard/search).

We've already run our fleet through dirctl import --type=a2a --dry-run — all 47 records transform and validate cleanly against OASF 1.1.0 (skills/domains enriched via the local extractor). We're ready to stand up a federated node and routing publish the fleet as soon as participation opens beyond the initial steering-committee phase.

What we'd bring to the testbed: a real production corpus of A2A-native agents with verifiable domain identity (happy to publish a JWKS at https://for.you.com/.well-known/jwks.json and sign records for name verification), plus feedback from the ARD-publisher side of the fence.

Who should we coordinate with to get onboarded? Thanks!

[muscariello]

@brainsparker we have recently opened the testbed to non TSC members, so you're welcome to participate.

You can start by looking at

https://dir.agntcy.org

There is a section dedicated to joining the testbed.

We have recently moved to Discord and you can find the link from the webpage I shared.

[brainsparker]

great thanks!