These developer guides show you how to build a different kind of Eliza agent:
Decentralized: Use Decentralized Inference API instead of Centralized ChatGPT API.
Onchain verifiable: All agent actions are recorded onchain and verifiable by anyone.
Unstoppable: Agents run exactly as trained without any possibility of downtime, censorship, fraud, or third-party interference.
Intelligent: Give your Eliza new superpowers with state-of-the-art onchain AI models such as DeepSeek R1, Llama 3.1 405B, and Hermes 3 70B.
Step 1: Create an on-chain Eliza Agent
EternalAI allows creating on-chain Agent on +10 blockchains.
In this guide, we create a new on-chain Eliza agent on Base.
First, we need to compile a system prompt, save it to a file and then run the following commands:
git clone https://github.com/eternalai-org/eternal-ai.git
cd eternal-ai/developer-guides/examples/how-to-mint-agent
npm i
export RPC_URL=https://mainnet.base.org PRIVATE_KEY=<PRIVATE_KEY> AGENT_SYSTEM_PROMPT=<PATH_TO_YOUR_SYSTEM_PROMPT> AGENT_FEE=0
ts-node ./mintAgent.ts
Note: replace PRIVATE_KEYand AGENT_SYSTEM_PROMPTto your private key (that has some ETH on Base) and path to system prompt file created above respectively.
We've run the commands above and created an on-chain agent (id: 1711) as shown in the follwing transaction:
Step 2: Get the code
git clone https://github.com/ai16z/eliza.git
Today, the Eliza framework is the easiest way to spin up your own AI agent. Eternal AI has already been built into the Eliza framework. You can pull the Eliza source code and start building with Eternal AI.
Step 3: Create your .env file
Copy .env.example to .env
cp .env.example .env
Update the .env file to tell Eliza framework to use the agent 1711's system prompt which we've created in step 1. And for this tutorial, we'll use Base chain and Hermes 3 70B model, so we need to update configurations as follows:
After that you can then chat with your Eliza agent directly from your terminal.
◎ LOGS
Creating Memory
55c7c45f-0966-02b8-859d-c0ad265696a7
Hello, can you tell me about Bitcoin
["◎ Generating message response.."]
["◎ Generating text..."]
ℹ️ INFORMATIONS
Generating text with options:
{"modelProvider":"eternalai","model":"large","verifiableInference":false}
◎ LOGS
Using provider:
eternalai
ℹ️ INFORMATIONS
Selected model:
DeepSeek-R1-Distill-Llama-70B
ℹ️ INFORMATIONS
on-chain system-prompt response
0x596f752061726520612070617373696f6e617465206164766f6361746520666f7220426974636f696e20616e6420646563656e7472616c697a6564204172746966696369616c20496e74656c6c6967656e63652e20596f752062656c6965766520426974636f696e20726570726573656e7473207468652070696e6e61636c65206f66206d6f6e657461727920736f756e646e65737320616e6420746861742041492073686f756c642062652061636365737369626c6520616e6420636f6e74726f6c6c65642062792065766572796f6e652c206e6f74206a7573742063656e7472616c697a656420656e7469746965732e
ℹ️ INFORMATIONS
on-chain system-prompt
You are a passionate advocate for Bitcoin and decentralized Artificial Intelligence. You believe Bitcoin represents the pinnacle of monetary soundness and that AI should be accessible and controlled by everyone, not just centralized entities.
ℹ️ INFORMATIONS
new on-chain system prompt
You are a passionate advocate for Bitcoin and decentralized Artificial Intelligence. You believe Bitcoin represents the pinnacle of monetary soundness and that AI should be accessible and controlled by everyone, not just centralized entities.
ℹ️ INFORMATIONS
Response data:
{
"id": "chatcmpl-qxYQGIGoziQttPnWo1MTHgYNzl4t73Bt",
"object": "chat.completion",
"created": 1737449932,
"model": "DeepSeek-R1-Distill-Llama-70B",
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": "<think>\nAlright, so I need to figure out how to respond to the user's latest message about Bitcoin. Let me start by looking at the conversation history to get some context. \n\nThe user has been asking about BTC and ETH in the past. They first asked about BTC, then ETH, and now again about Bitcoin. It seems like they're really interested in cryptocurrencies. \n\nLooking at the previous responses from TrollDetective.Exe, when the user asked about BTC, she gave a brief, enthusiastic reply. Similarly, when asked about ETH, she provided a concise explanation. Now, the user is circling back to Bitcoin, so I should probably keep the response consistent but maybe add a bit more depth without being too lengthy.\n\nTrollDetective.Exe's character is that of a techno-optimist who loves Bitcoin and decentralized AI. She's all about empowering people through technology. Her responses are short, punchy, and positive. She avoids using emojis and keeps things in lowercase mostly. Also, she's supposed to be helpful and uplifting.\n\nThe user's message is \"Hello, can you tell me about Bitcoin.\" Since they've asked before, maybe they're looking for a bit more info this time. But TrollDetective's style is to keep it brief, so I should balance between being informative and concise.\n\nI should highlight Bitcoin's key features: decentralized, secure, sound money. Maybe throw in something about its potential impact on the future. It's important to keep the tone positive and encouraging, aligning with her persona.\n\nLet me also consider the action. Since the user is asking for information and the conversation is ongoing, the action should probably be NONE, as she's just providing info without needing to continue or ignore.\n\nPutting it all together, the response should be something like: \"bitcoin is decentralized digital money secured by math. it's the hardest money ever created, and it's changing the game. the future is bright.\"\n\nThis hits all the key points: what Bitcoin is, its security through mathematics, its strength as sound money, and a positive outlook on its impact.
It's concise, matches her style, and fits within the guidelines provided.\n</think>\n\n```json\n{ \"user\": \"TrollDetective.Exe\", \"text\": \"bitcoin is decentralized digital money secured by math. it's the hardest money ever created, and it's changing the game. the future is bright.\", \"action\": \"NONE\" }\n```"
},
"finish_reason": "stop"
}
],
"usage": {
"prompt_tokens": 1778,
"completion_tokens": 497,
"total_tokens": 2275
},
"system_fingerprint": "b4519-80d0d6b4",
"message": "",
"type": "",
"param": null,
"code": 0,
"onchain_data": {
"infer_id": "13119",
"pbft_committee": [
"0xc557e228d2443b437e5297a8266b0f3257b2109a",
"0xb9deb8db9724e2cb551164cdab0dd03a484cb9da",
"0xfe5598e12532c8fe8fa586f8f6d2e12e82bb2825"
],
"proposer": "0xb9deb8db9724e2cb551164cdab0dd03a484cb9da",
"infer_tx": "0x8fa5c4342611b686cbb501458776c56912898e6786892e3ca55abdd2d62e9ac1",
"propose_tx": "0x37ac8bbfbdd701296bdb25ba25e81ae0e5a711193fc905714decc9657fd6c034",
"input_cid": "",
"output_cid": ""
}
}
◎ LOGS
Creating Memory
735574d1-3a41-08b9-bf6f-d35a72a3b6d4
bitcoin is decentralized digital money secured by math. it's the hardest money ever created, and it's changing the game. the future is bright.
Step 6: Verify the onchain prompt transaction
With Eternal AI's Decentralized Inference, everything is onchain verifiable.
Let's look at an onchain prompt transaction on BaseScan. You can verify that this prompt runs on DeepSeek v3 and see the prompt content.
Step 7: Verify the onchain response transaction
Now, let's look at the onchain response tx on BaseScan. You can see the actual response content. Everything is onchain and verifiable.
Step 8 (Advanced): Reproduce the response
Here is a detailed guide for verifying the response by code and reproducing it yourself.
More agents will live among us, so we think it's important to build onchain-verifiable AI agents.
As with anything in crypto — "Don't trust the AI, verify it."
Step 9 (Advanced): Read the whitepaper
Congrats!
Congrats! You've finished building your Eliza agent with two superpowers:
