Official Community Partner of

No Slides. No Pitches. Just Run Loops.

The Agent Jam is an unconference-style gathering built specifically for AI engineers, builders, and product strategists who prize software craftsmanship over ad-hoc "vibe coding."

We are leaving the high-level marketing decks at home. If you want to take the floor, you must bring an active terminal, an open IDE, a live prompt workflow, or a working multi-agent setup.

How It Works

An unconference is entirely peer-driven. We do not curate a rigid speaker lineup beforehand. Instead, the agenda is built on-the-fly by the people in the room during the kickoff.

• The Whiteboard Grid: At 10:00 AM, a massive physical grid goes up on the wall.

• Claim Your Slot: Anyone—whether an expert developer or a sharp product builder—can pitch a hack and claim a strict 15-minute slot on the grid.

• 12 + 3 Bounding: Every session is strictly capped at 12 minutes of live demonstration and 3 minutes of peer troubleshooting and Q&A.

• Show the Plumb Lines: Don't just show the shiny frontend output. Show the logs, the system prompt, the custom API payload, or the logic graph that powers it.

What We Are Jamming On

While the agenda is fluid, the collective focus centers on real-world engineering discipline and architecture:

• Autonomous Agents & Multi-Agent Systems: Coordination frameworks, custom run loops, and task orchestration.

• Model Context Protocol (MCP): Building and connecting custom MCP servers to close the gap between LLMs and local development environments.

• Local Implementations: Running open models entirely on local hardware configurations, utilizing tools like Ollama and local agent orchestration infrastructure.

• Production-Grade Workflows: Advanced prompt chaining, rigid validation steps, evaluation guardrails, and deterministic outputs from non-deterministic models.

The Sandbox Guardrails

To keep the technical quality exceptionally sharp, all participants must agree to three simple rules:

1. Fallbacks are Mandatory

If your local model or cloud environment takes longer than 3 minutes to spin up or fix an unexpected bug, you must have a backup recording ready. The clock does not pause for infrastructure configuration issues.

2. No Idea is Too Small

A robust bash script utilizing an LLM CLI tool to clean messy production logs is vastly more valuable to this community than a broken, over-engineered autonomous agent. Show what works.

3. Peer Critique, Not Judgement

This is a collaborative sandbox. We are here to teardown systems, expose edge cases, and collectively figure out how to build reliable AI infrastructure.

Getting Here

• Location: Kovan Labs, Second Floor, IndiQube Echo, SITRA, Opposite to passport office, Coimbatore.

• Timing: Saturday, May 30th, 2026. Doors open at 10:00 AM IST for grid mapping and coffee.

• Prerequisites: Bring your laptop, charger, and an active local environment. High-density power and high-speed Wi-Fi will be provided.