@user Can the submission video be longer than 3 minutes?

While using a2a chat response from archestra how do I change the model? For agent chat in archestra UI I am using gemini-2.5-flash but while calling the a2a endpoint from my backend it defaults to deep-research-pro-preview . How do I change it?

🛡️ Archestra Sentinel: Real-Time Governance for AI Agents

Just submitted my project for the 2 Fast 2 MCP Hackathon! 🚀

Sentinel is a security layer that bridges the gap between the host and the admin.

How I used Archestra: I utilized the Archestra Gateway Mode and a secure TCP Bridge to expose the host’s Docker daemon. This allows Sentinel to pull real-time container metrics for our Trust Intelligence Engine—evaluating agents based on Identity, Privilege, Network, and Resources—all while maintaining a secure, governed environment.

• Active Governance: One-click Kill/Quarantine for unauthorized agents.

• Cost Tracking: Live infrastructure burn-rate analytics.

• Trust Score: 4-vector logic (Identity, Privilege, Network, Resources).

Demo: https://www.youtube.com/watch?v=DppGd74_5ZE

GitHub: https://github.com/mohitjeswani01/archestra-sentinel

As a final-year student, this project was an incredible journey into AI observability. I’m now looking to scale these architectural patterns into complex, real-world systems and connect with teams building the future of AI infrastructure. :clinking_glasses:.

Hi, this is really important

Ill really appreciate if someone from the team responds

Is it ok for the video to be more than 3 minutes???

FlowForge is an AI-powered no-code backend builder that converts natural language into production-ready APIs within seconds. Users describe their requirements, and the platform generates complete workflows with validation, database operations, authentication, and structured responses. It features a visual ReactFlow editor, real-time execution logs, chat-based intelligent workflow mutations, and a powerful Deploy to Archestra feature for instant production hosting. Built using Next.js, Express, MongoDB, WebSockets, and a multi-agent architecture (SupervisorAgent, BuilderAgent, MCPRuntime), FlowForge reduces backend development time from days to under a minute while maintaining transparency and scalability.

Deployed Link : https://flowforge-i871.vercel.app/landing

Youtube Link : https://youtu.be/VcgeOuzxbdg

Hey everyone, here's my submission for 2 Fast 2 MCP 🏁

MCP Guardian — a security auditor that scans MCP servers for vulnerabilities and auto-enforces policies through Archestra.

It detects prompt injection, data exfiltration, command injection, the "Lethal Trifecta" (read + write + execute on a single server), and writes blocking policies directly back to Archestra's policy engine. Full closed loop: scan → detect → enforce.

6 Archestra API endpoints used, Terraform provider for deployment, and a malicious demo server included to showcase detection live.

Demo: https://guardian.aryankumar.dev

GitHub: https://github.com/aryan877/mcp-guardian

Video: https://youtu.be/0YnpyLxh4zo

Hey team, while using archestra on mobile, i faced a lot of issues related to UI. So should I open issue on GitHub or you guys are already working on responsiveness of the site?

Hi will we get any certificate of participation? If yes, then when?

Community call is tomorrow! See you!

Propose your agenda in 🧵

Hey everyone, I'm pretty late to this, I had submitted the google form on time but wanted to wait to post this in the channel.

I wanted to create this for a long time, and this couldn't be more of a perfect opportunity to get this done.

I had a feeling since a long time, that every other AI-related SaaS that comes out now-a-days is either helping the tech people, or people who know the very fancy side of AI, like all of us.

But what about the people who are not exposed to it like we are?

Don't they deserve to use it? Absolutely, they do.

AI should help in increasing everyone's productivity and business, not just of founders and techies.

Hence the 2 fast 2 MCP hackathon felt like a perfect time to find an idea for this thought of mine.

• So I have launched Preffer, it makes everyone prefer you!

• Live at : https://preffer.me (uses archestra's deployed backend- https://archestra.preffer.me, mcp url: https://app.preffer.me/mcp - the api key can be generated via the credentials page on the preffer dashboard, exposes 27 tools.)

• Preffer gives anyone a professional page (AI helps in building it) plus an AI representative that talks to visitors, answers questions, captures intent, and turns conversations into business outcomes. Your AI representative gets it's name, personality and instructions from you and is reachable by a public URL that Preffer provides you.

• Comes with crazy analytics too!

• One feature I was especially excited about was enabling AI representatives (which are Archestra agents 🙂) to reach out to users they’ve interacted with, if needed, to reduce the hassle for the page owner. I implemented this in Archestra using the existing components which were super helpful. - https://github.com/archestra-ai/archestra/pull/2820

Youtube demo URL: https://youtu.be/tsDKUfw1SX8?si=BmpODX64Dukp22qA

Please use it yourself, giving 500 free credits! Any feedback is welcome!!

Thanks for this opportunity, Archestra!!!! Couldn't be more thankful.

Hey, no community call today?

New release! 🚀🚀🚀

Bug Fixes

Miscellaneous Chores