LiveKit Hackathon Starter Kits

LiveKit at the Conversational AI Hackathon

Build a voice AI agent with LiveKit Agents and Moss real-time semantic search. Hosted by Moss (F25) at Y Combinator, San Francisco on June 6–7, 2026. Everything below is what you need to ship in a weekend.

Voice models are cheap and fast, retrieval is the bottleneck. With Moss, real-time semantic search is solved: agents hold a fluid conversation, instantly look up complex facts, and remember across turns. This starter wires that into a LiveKit voice agent so you can focus on your idea.

1. Install the LiveKit Docs MCP server

The fastest way to keep your coding agent on the rails. It pulls live docs, code examples, and changelogs straight into Cursor / Claude Code / VS Code so you stop fighting hallucinations.

Claude Code

claude mcp add --transport http livekit-docs https://docs.livekit.io/mcp

Cursor — one-click install

VS Code

code --add-mcp '{"name":"livekit-docs","type":"http","url":"https://docs.livekit.io/mcp"}'

Codex

codex mcp add --url https://docs.livekit.io/mcp livekit-docs

Full reference and other clients (Gemini CLI, Antigravity, Copilot CLI): docs.livekit.io/mcp.

Or use the LiveKit Agents skill

If you'd rather use a Claude Code skill instead of (or alongside) the MCP server:

npx skills add https://github.com/livekit/agent-skills --skill livekit-agents

Activates automatically for relevant tasks, or invoke with /livekit-agents.

2. Clone the Moss starter kit

git clone https://github.com/livekit-examples/moss-hacker-starter

A LiveKit voice agent (Python) wired into Moss semantic search with a Next.js frontend. Demonstrates the patterns you'll build on:

  • Semantic search (RAG) — a search_knowledge tool runs sub-10ms Moss queries over your knowledge index
  • Agentic memory — remember_fact / recall_facts store and recall per-user details across the conversation
  • Live "Knowledge Matches" panel — see the retrieved chunks + relevance scores in the UI in real time
  • LiveKit Inference — STT/LLM/TTS by model name, so the only secrets you wire are LiveKit + Moss

You'll need two free accounts:

  • LiveKit Cloud — sign up at cloud.livekit.io and create a project to get your LIVEKIT_URL, LIVEKIT_API_KEY, and LIVEKIT_API_SECRET (grab your $50 in credits in step 3)
  • Moss — get your MOSS_PROJECT_ID / MOSS_PROJECT_KEY from portal.usemoss.dev

Then follow the README:

pnpm setup        # install agent + frontend
# add your LiveKit keys — `lk app env` writes them once the CLI is signed in (step 4),
# or paste LIVEKIT_URL / LIVEKIT_API_KEY / LIVEKIT_API_SECRET into agent-py/.env.local and frontend/.env.local
# paste MOSS_PROJECT_ID / MOSS_PROJECT_KEY into agent-py/.env.local
pnpm moss:index   # build the knowledge + memory indexes
pnpm dev          # run the agent + frontend together

Swap agent-py/knowledge.json for your own docs, re-run pnpm moss:index, and you're searching your own content.

3. Redeem your $50 in inference credits

All attendees get a 7-day free trial of the LiveKit Ship plan — $50 in inference credits, no credit card required.

  1. Create a project at cloud.livekit.io
  2. Redeem at cloud.livekit.io/projects/p_/redeem
  3. Code: HACK-MOSS-YC

4. Ship it

Install the LiveKit CLI (one-time):

brew install livekit-cli
# or
curl -sSL https://get.livekit.io/cli | bash

CLI setup and reference: docs.livekit.io/intro/basics/cli.

When your agent works locally, deploy from inside the agent directory:

lk agent create

Then test against the frontend you cloned.

Tracks

Pick a track and make the conversation feel effortless:

  • Lead Gen — agents that nurture and convert inbound leads
  • Support — customer-service bots that instantly pull docs and user history
  • Co-Pilot — ambient agents that listen in and display live context

The starter's RAG + agentic-memory patterns map directly onto all three — especially Support, where the live Knowledge Matches panel shows the agent grounding every answer.

Reference