1. The Thesis (what to build)

Build a memecoin research agent powered by Grok that takes a token's contract address (or ticker) and returns a sourced research report. The agent has to research the token itself, deciding which tools to call and in what order. A single prompt pasted into a chat window doesn't count.

Every agent must cover three sources:

Source	How	What to look at
The timeline	Grok's X Search tool	Who's posting about it, when the posts started, whether accounts look coordinated or bought, KOL mentions, and the dev's own posting history
The chain	Function calling into your own onchain tools (RPC, indexers, explorers)	Mint and freeze authority, top-holder concentration, dev and team wallets, bundled or sniped launch buys, liquidity depth and whether it's locked, where the funding came from
The web	Grok's Web Search tool	Website, docs, earlier projects by the same team, any rug or scam reports

The report should then check the three sources against each other. The most useful findings come from gaps between them, such as a token trending on X while three wallets hold 60% of the supply, or a "community takeover" where the dev is still quietly selling.

Build on any chain (Solana, Base or others), with any stack and any frontend: a web app, Telegram or Discord bot, CLI or browser extension. Some starting ideas:

• Contract address → report: paste an address and get a one-page report with a risk score, where every claim links to its source.
• Timeline vs. chain: a split view showing X hype on one side and onchain reality on the other, with the mismatches highlighted.
• Dev dossier: follow the deployer wallet and its X handle across earlier launches and show how those went.
• Watchlist agent: watches new launches and flags patterns like bundled launches, fake engagement and unlocked liquidity as they happen.
• Reply guy: a bot that answers "@agent is this legit?" with a short report it has actually researched.

2. Due Diligence (requirements)

• Grok is the brain. The agent must run on a Grok model through the xAI API, and use X Search plus at least one onchain function call. Web Search, Code Execution, Structured Outputs and Collections are encouraged.
• Every claim has a receipt. Each finding in the report links to its source: an X post, a transaction, a wallet, or a web page. Grok returns source URLs for its built-in tools, so show them.
• Structured output. The report follows a fixed format (for example with Structured Outputs) so reports can be compared across tokens.
• Research, not advice. The agent reports findings and red flags. It doesn't say "buy," "sell" or give price targets, and every report carries a not-financial-advice notice.
• Read-only. The agent doesn't trade, sign transactions or hold funds.

3. The Receipts (submission)

You have 4 days. Your submission needs:

• A public GitHub repo with a README covering how to run it, which Grok model and tools it uses, and every onchain data source it calls.
• The backtest: run your agent on at least 10 real tokens that already launched, some that rugged and some that survived, and include the reports. Show what it flagged before the outcome, and be honest about what it missed.
• A demo video (YouTube, Loom, Vimeo or X) of 3 minutes or less that shows a contract address going in and a sourced report coming out, with the agent's tool calls visible.
• A short description of what you built and why.
• No API keys in the repo. Use environment variables and include a .env.example file.
• One submission per person; you can update it until submissions close.

Solo builders and teams of up to 4. Anything started before the hackathon opened must be clearly marked, along with a description of the new work.

4. The Verdict (judging)

• Voting opens the moment submissions close and runs for 24 hours.
• Token holders vote, weighted by how much they hold at the snapshot (and still hold when they vote).
• The submission with the most vote weight wins. Ties go to the earlier submission.

5. The Payout (prize)

• The winner receives the entire prize pool (sponsorship + 90% of creator fees earned so far).
• Every future fee claim is forwarded to the winner's payout address, forever.

6. Red Flags (disqualification)

Plagiarised work, non-functional demos or malicious code are removed from voting. So are:

• agents that don't actually use Grok;
• reports with made-up or unsourced findings;
• agents that shill tokens, make buy or sell calls, or promote a token the builder holds;
• bots that spam replies or mass-tag accounts on X;
• anything that trades, signs transactions or takes custody of funds;
• any repo that exposes API keys, private keys or seed phrases.

---

Resources

Grok / xAI API

• Quickstart
• Models · Pricing
• Tools overview
• X Search
• Web Search
• Function calling
• Code execution
• Structured outputs
• Responses API reference
• Full docs index (llms.txt)

Onchain data (Solana)

• RPC overview and cluster endpoints
• RPC HTTP methods
• RPC WebSocket methods
• Choosing an RPC provider