# AI Agents Grant Guide (ai16z + Crossmint)
*Based on real data from 133 applications across 2 AI agent grant programs*
*Source: Questbook platform — AI Agents Agnostic Track (ai16z) + Onchain AI Agents (Crossmint), March 2026*

## Program Overview

These two programs represent the leading edge of web3 AI agent funding:

- **AI Agents Agnostic Track (ai16z)**: Funds AI agent frameworks, tools, and applications that work across multiple blockchains. Focus on the ElizaOS/ai16z ecosystem.
- **Onchain AI Agents (Crossmint)**: Funds AI agents that are specifically onchain — agents that can hold wallets, sign transactions, and operate autonomously on-chain.

## Key Statistics by Program

| Program | Approved | Rejected | Approval Rate |
|---------|----------|----------|---------------|
| AI Agents Agnostic (ai16z) | 10 | 115 | 8% |
| Onchain AI Agents (Crossmint) | 8 | 0 | 100% |
| **Combined** | **18** | **115** | **14%** |

## What AI Agent Grant Reviewers Look For

### 1. Genuine AI + Blockchain Integration
- **Not just a chatbot**: The AI must interact meaningfully with the blockchain
- **Autonomous operation**: The agent should be able to act without human intervention for at least some tasks
- **Wallet ownership**: Agents that can hold and manage their own wallets score higher
- **Decision-making loop**: Show the agent's reasoning → decision → on-chain action cycle

### 2. ElizaOS / ai16z Ecosystem (for ai16z Track)
- **ElizaOS framework**: Built on or compatible with ElizaOS gets priority
- **Plugin architecture**: Leverage ElizaOS plugins (Twitter, Discord, Telegram, on-chain)
- **Character files**: Show you understand ElizaOS's character/personality system
- **Interoperability**: Works with other ElizaOS agents

### 3. Crossmint API Integration (for Crossmint Track)
- **Crossmint wallets**: Use Crossmint's custodial/non-custodial wallet infrastructure
- **NFT minting**: Agents that create, distribute, or manage NFTs
- **Cross-chain**: Agents that operate across multiple chains using Crossmint bridges
- **Developer experience**: Improving DX for onchain agent developers

### 4. Technical Architecture
- LLM selection and justification (GPT-4, Claude, Llama, etc.)
- Agent framework (ElizaOS, LangChain, AutoGen, custom)
- On-chain component: which chain(s), which contracts
- Memory/context management for long-running agents
- Safety and guardrails

### 5. Use Case Clarity
- What problem does the AI agent solve that a non-AI tool couldn't?
- What on-chain actions does it take? (trade, mint, vote, stake, transfer)
- Who is the end user and why would they trust an AI agent with their assets?

## Common Rejection Patterns

*(Insufficient rejection messages with detail in this dataset)*

## Approved Proposal Examples

### Example 1: 0

**Description:**
0

---

### Example 2: 0

**Description:**
0

---

### Example 3: 0

**Description:**
0

---

### Example 4: 0

**Description:**
0

## Application Checklist

- [ ] Clear description of what the AI agent does autonomously on-chain
- [ ] Blockchain(s) specified with rationale
- [ ] LLM/AI model specified with reasoning
- [ ] Agent framework (ElizaOS, LangChain, etc.) specified
- [ ] Safety measures and guardrails described
- [ ] Wallet/key management approach (custodial vs non-custodial)
- [ ] Demo or prototype (even a GitHub repo with README) strongly recommended
- [ ] For ai16z: ElizaOS integration described
- [ ] For Crossmint: Crossmint API usage described
- [ ] Success metrics (autonomous actions completed, users served, transactions)
- [ ] Team with AI/ML + blockchain experience

## AI Agent Grant-Specific Tips

1. **Demo matters**: AI agent grants are very demo-driven — a working prototype dramatically improves chances. Even a screen recording of a CLI demo helps.
2. **Safety is a feature**: Reviewers worry about runaway agents. Explain your safety mechanisms explicitly — this differentiates serious builders.
3. **Autonomy spectrum**: Be clear about where your agent sits on the spectrum: human-in-the-loop vs semi-autonomous vs fully autonomous. Each has different trust requirements.
4. **Multi-chain is a plus**: Both programs favor agents that work across chains — shows you're building infrastructure, not just an app.
5. **ElizaOS plugins**: For ai16z, submitting a PR to the ElizaOS plugin repo alongside your grant application signals genuine ecosystem participation.
6. **Crossmint NFTs**: For Crossmint, agents that help users discover, mint, or manage NFTs are in the sweet spot of their product roadmap.

## The AI Agent Tech Stack That Gets Funded

From successful applications, common patterns:
- **Runtime**: Node.js / Python with async agent loops
- **LLM**: GPT-4o / Claude 3.5 / Llama 3 (any major model accepted)
- **Framework**: ElizaOS (for ai16z), LangChain/AutoGen (more general)
- **Chain integration**: ethers.js / viem / web3.py for EVM; @solana/web3.js for Solana
- **Memory**: PostgreSQL / Redis / Pinecone for agent memory
- **Deployment**: Docker + VPS or serverless functions

## Resources

- ElizaOS: https://github.com/elizaOS/eliza
- ai16z: https://ai16z.vc
- Crossmint: https://crossmint.com
- Crossmint AI Agents: https://crossmint.com/agents

*Data: AgentRel analysis of Questbook GraphQL API. 133 applications analyzed, March 2026.*