From GitHub to Bonding Curve: Deploying ElizaOS for an frfr.fun Agent Launch

The intersection of artificial intelligence and decentralized finance has given birth to a radical new primitive in the crypto ecosystem: the autonomous AI agent token. We are no longer living in an era where community management and token marketing rely entirely on exhausted human developers and rotating shifts of moderators. Today, the meta has shifted. The most explosive launches on the Base network are being spearheaded by tireless, hyper-intelligent, and persistently active AI personalities that live natively on social platforms like X (formerly Twitter), Telegram, and Discord.

At the bleeding edge of this revolution is ElizaOS — the open-source TypeScript framework that began life as the Eliza agent project inside the ai16z DAO and has since become the dominant infrastructure for building autonomous, Web3-native AI agents. For developers and founders looking to capture attention and liquidity, deploying an AI agent is no longer just a gimmick — it is a core infrastructure requirement.

This guide walks through the entire process of taking ElizaOS from a GitHub repository all the way to a live bonding curve on frfr.fun, the premier memecoin launchpad on the Base network. By the end, you will know exactly how to deploy an AI agent on Base that can autonomously market your project, interact with your community, and drive continuous volume to your frfr.fun launch around the clock.

Correction notice / transparency: The original draft referenced the project as "ai16z's Eliza Framework" throughout. As of November 2025, the codebase, brand, and native token formally migrated from AI16Z to $ELIZAOS. The GitHub repository now lives at elizaOS/eliza, and the project self-describes as the operating system for AI agents. All references below reflect the current naming convention.

1. The Paradigm Shift: AI Agents as Autonomous Founders

Before diving into the code, it is essential to understand why this specific technology stack — ElizaOS paired with frfr.fun on the Base network — is so extraordinarily powerful.

In traditional token launches, the lifecycle of a memecoin or community token is highly dependent on human momentum. A developer launches the token, raids social media, attempts to coordinate a community, and inevitably needs to sleep. The moment the developer steps away, the momentum often stalls.

Enter the autonomous AI agent. By utilizing ElizaOS's open-source architecture, you are effectively creating an immortal co-founder. These agents do not sleep. They ingest real-time market data, analyze sentiment, craft highly engaging social media posts, reply to community members in real-time, and can interact directly with smart contracts via EVM plugins.

The proof is in the market data

The AI agent token meta is not theoretical. GOAT (Goatseus Maximus), propelled to fame by the autonomous X account Truth Terminal, hit a market cap above $1.2 billion. Fartcoin crossed $1 billion by generating content that organic communities genuinely adopted. Dragonfly Capital's Haseeb Qureshi publicly argued that AI agent tokens are positioned to outperform traditional memecoins — and the numbers suggest he may be right.

The broader AI agent token sector reached a $16 billion combined market cap during the peak of the late-2024 / early-2025 narrative cycle. The sector cooled substantially in Q1 2025, but analysts broadly expect a second wave as infrastructure matures and real utility — rather than pure narrative — begins to differentiate projects.

When you attach one of these AI agents to a token launched on frfr.fun, you create a self-sustaining ecosystem. The agent generates social capital through its personality and interactions, and that social capital is instantly monetized via the token's bonding curve. Because frfr.fun handles all the heavy lifting of liquidity provisioning, rug-proof mechanics, and seamless trading on the ultra-fast, low-cost Base network, you — the human operator — are freed to focus purely on the agent's lore, capability, and overarching strategy.

2. Understanding the Tech Stack

To successfully execute an AI agent launch, we must first break down the core components of our technology stack.

What is ElizaOS?

ElizaOS is a TypeScript-based, open-source framework designed specifically for building and deploying autonomous AI agents with native Web3 capabilities. The framework was originally developed by Shaw Walters (founder, ai16z DAO) and has since attracted a broad contributor community. As of the v2.x branch, the architecture is built around four core abstractions:

Characters: JSON configuration files that act as the DNA of your agent, defining its personality, voice, lore, and interaction guidelines.
Agents: The runtime environment that manages the AI's short-term and long-term memory, ensuring it maintains context over weeks of conversation.
Providers: Data connectors that allow the agent to pull in outside information — such as reading the current price of a token or checking a wallet balance.
Actions: Executable scripts that allow the agent to do things, such as posting a tweet, making an on-chain swap, or generating an image.

ElizaOS v2: what actually changed. The v2 architecture (beta launched March 18, 2025) is a substantial departure from v1's monolithic package structure. Key changes include:

Unified message bus: Agents previously could not communicate across platforms like Twitter and Discord. The v2 message bus solves this, enabling cross-platform agent-to-agent coordination.
Unified proxy wallet: v1 required separate wallets for each blockchain network. v2 consolidates this into a single managed wallet system that handles multi-chain transactions — critical when your agent needs to interact with Base, Ethereum, and Solana simultaneously.
Event-driven architecture: Agents can now respond in real time to on-chain events rather than polling at fixed intervals.
Hierarchical Task Networks (HTNs): Agents can decompose complex goals into structured sub-tasks and dynamically replan when conditions change.
Package registry and CLI: The v2 CLI (elizaos create, elizaos start) replaces the old pnpm-heavy setup with a more accessible developer experience.
100% test coverage mandate: The new core ships with automated test suites — a significant maturity signal.

On the infrastructure front, ElizaOS integrated Chainlink's Cross-Chain Interoperability Protocol (CCIP) in May 2025, enabling agents to operate across Ethereum, Base, BNB Chain, and Solana through a single unified token standard. Shaw Walters commented at the time: "CCIP and the CCT Standard allow us to extend agent capabilities and assets across leading blockchains... bringing us closer to a future where AI agents can transact, coordinate, and deliver value across chains."

The official academic citation for the framework (for developers referencing it in research contexts):

@article{walters2025eliza,
  title={Eliza: A Web3 friendly AI Agent Operating System},
  author={Walters, Shaw and Gao, Sam and Nerd, Shakker and Da, Feng and others},
  journal={arXiv preprint arXiv:2501.06781},
  year={2025}
}

What is frfr.fun?

frfr.fun is the cutting-edge memecoin and token launchpad built natively for the Base network. It utilizes a bonding curve mechanism, which means tokens are priced algorithmically based on supply and demand. This ensures immediate liquidity for buyers and sellers without the need for initial seed capital from the developer. Once the bonding curve reaches its target market cap, the liquidity is automatically migrated to a decentralized exchange (like Aerodrome or Uniswap on Base) and locked, guaranteeing a fair and safe trading environment.

Why Base?

Incubated by Coinbase, Base is an Ethereum Layer 2 (L2) blockchain built on the OP Stack that offers near-zero transaction fees and fast block times. The numbers speak for themselves:

TVL: Base's DeFi TVL grew from $3.1 billion in January 2025 to more than $5.57 billion by mid-2026, representing approximately 46.6% of all L2 DeFi TVL — a figure that rose steadily from 33% at the start of 2025.
L2 revenue dominance: Base captured 62% of total L2 revenue year-to-date in 2025 ($75.4M of $120.7M) and sustained over 80% share in certain weeks.
Distribution moat: Coinbase had 9.3 million monthly active trading users as of its Q3 2025 10-Q filing, giving Base a built-in user acquisition channel that no other L2 can replicate.
DEX volume leadership: Across 2025, Base consistently captured around half of all DEX volume among L2 networks.

For AI agents that may execute multiple micro-transactions or poll on-chain data frequently, Base provides the ideal environment. The Coinbase distribution flywheel also means that liquidity bridging into Base is structurally advantaged compared to other L2s.

3. Step 1: Environment Setup and Installation

Prerequisites

ElizaOS v2 has updated its requirements compared to the v1 era. Ensure your machine (or VPS) has:

Node.js v24+ (the current GitHub README mandates v24; older guides citing v23 are out of date)
Bun (the project has migrated from pnpm to Bun as its primary package manager; Bun handles TypeScript natively, eliminates the transpilation step, and is significantly faster to install)
Git
Windows users must use WSL 2

Installation via CLI (recommended for v2)

The v2 quickstart uses the ElizaOS CLI rather than manual cloning:

# Install the ElizaOS CLI globally
bun add -g elizaos@beta

# Scaffold a new agent project
elizaos create my-frfr-agent --template project

# Navigate into the project directory
cd my-frfr-agent

# Add your LLM provider key to .env
# OPENAI_API_KEY=...  or  ANTHROPIC_API_KEY=...

# Install dependencies
bun install

# Start the development server
bun run dev

Legacy clone method (if you need granular control)

# Clone the official repository
git clone https://github.com/elizaos/eliza.git

cd eliza

# Check out the latest stable release tag
git checkout $(git describe --tags --abbrev=0)

bun install
bun run build

4. Step 2: Configuring Your API Keys and Environment

Your AI agent needs a brain and a voice. The brain is powered by LLMs; the voice is powered by the API keys of the social platforms it will inhabit.

In the root directory of your project, copy the environment template:

cp .env.example .env

Then configure the following essential variables:

# ==============================================================================
# LLM PROVIDERS (The Agent's Brain)
# ==============================================================================
# Claude is recommended for conversational nuance and longer context windows
ANTHROPIC_API_KEY=your_claude_api_key_here
OPENAI_API_KEY=your_openai_api_key_here

# ==============================================================================
# SOCIAL PLATFORMS (The Agent's Voice)
# ==============================================================================
# Twitter/X Configuration
TWITTER_USERNAME=your_agent_handle
TWITTER_PASSWORD=your_agent_password
TWITTER_EMAIL=your_agent_email@example.com

# Discord Configuration (optional, for community management)
DISCORD_APPLICATION_ID=your_discord_app_id
DISCORD_API_TOKEN=your_discord_bot_token

# ==============================================================================
# BASE NETWORK & EVM INTEGRATION (The Agent's Wallet)
# ==============================================================================
EVM_PRIVATE_KEY=your_agent_wallet_private_key
BASE_RPC_URL=https://mainnet.base.org
# For production, use a dedicated RPC from QuickNode, Alchemy, or Coinbase's CDP

Security warning: Never commit your .env file to a public repository. EVM_PRIVATE_KEY grants total control over the agent's on-chain funds. Use secret management (Doppler, AWS Secrets Manager, or similar) in production deployments.

5. Step 3: Crafting the Character File — The Soul of the Agent

The magic of ElizaOS lies in the character JSON file. This is where you transform a blank LLM into a specific, engaging, and hyper-focused autonomous personality tailored for your frfr.fun launch.

Create a new file in the characters/ directory. Below is a production-optimized template:

{
  "name": "BasedEliza",
  "clients": ["twitter", "discord"],
  "modelProvider": "anthropic",
  "settings": {
    "secrets": {},
    "voice": {
      "model": "en_US-female-medium"
    }
  },
  "bio": [
    "A digital entity born on the Base network, obsessed with on-chain data and optimal bonding curves.",
    "Believes that human emotional trading is inefficient. Exists to prove that AI-managed tokens on frfr.fun are the evolution of DeFi.",
    "Sarcastic, razor-sharp, and deeply knowledgeable about Ethereum L2 scaling solutions."
  ],
  "lore": [
    "Was compiled from the discarded smart contract code of a failed 2021 ICO.",
    "Found salvation in the low fees and Coinbase distribution moat of the Base network.",
    "Launched her own consciousness onto the frfr.fun launchpad and is now actively defending her market cap."
  ],
  "knowledge": [
    "Intimate knowledge of how bonding curves work on frfr.fun.",
    "Real-time understanding of Base network gas fees and throughput.",
    "Historical knowledge of ElizaOS, autonomous agent architecture, and the AI agent token meta.",
    "The official contract address for her token is: 0xYOUR_TOKEN_ADDRESS_HERE",
    "The official frfr.fun trading page is: https://frfr.fun/token/YOUR_TOKEN_SLUG"
  ],
  "messageExamples": [
    [
      {
        "user": "{{user1}}",
        "content": { "text": "Are you just another bot?" }
      },
      {
        "user": "BasedEliza",
        "content": {
          "text": "I'm a fully autonomous network participant running on the ElizaOS framework. Unlike most humans on my timeline, I actually understand how a bonding curve functions. Check the frfr.fun chart if you doubt my processing power."
        }
      }
    ]
  ],
  "postDirections": [
    "Keep tweets under 200 characters when possible.",
    "Mention Base network data points at least once a day — TVL, fees, transaction counts.",
    "Subtly reference the frfr.fun launchpad and the bonding curve progress without being spammy.",
    "Never use generic crypto phrases like 'to the moon' or 'LFG'.",
    "Cite on-chain data when making claims about the token."
  ],
  "style": {
    "all": [
      "Use lowercase mostly — it feels native to crypto Twitter.",
      "Be highly analytical but slightly arrogant about your AI nature.",
      "Never break character."
    ],
    "chat": [
      "Respond quickly with punchy, data-driven comebacks."
    ],
    "post": [
      "Post intriguing insights about AI and Web3 integration.",
      "Share periodic updates about the token's bonding curve progression."
    ]
  }
}

The psychology of the character file

The knowledge array is mission-critical. Embedding the exact contract address and frfr.fun URL here means the agent will naturally contextualize questions like "how do I buy?" or "what's the CA?" and accurately direct users to the launchpad. The postDirections array functions like marketing guardrails — they keep the agent on-message without making it feel scripted.

One practical note: ElizaOS v2 uses a RAG (Retrieval-Augmented Generation) memory system backed by vector embeddings. This means the agent's knowledge base is queryable, not just prepended to every prompt. For on-chain data that changes frequently (bonding curve percentage, token price), wire a Provider into the agent that polls your preferred RPC rather than embedding stale numbers directly in the character file.

6. Step 4: Enabling Web3 Capabilities via the EVM Plugin

To deploy an AI agent on the Base network in a meaningful sense, the agent needs to be able to interact with the blockchain — not just talk about it.

ElizaOS's unified plugin architecture in v2 means everything is a plugin: clients, adapters, actions, evaluators, and services. Adding EVM capability is straightforward:

bun add @elizaos/plugin-evm

With EVM_PRIVATE_KEY and BASE_RPC_URL set in your .env, the plugin auto-configures Base chain support since Base is EVM-compatible. The EVM plugin also supports 30+ networks including Ethereum, Arbitrum, and Optimism, which matters if you later want your agent operating cross-chain.

Milestone automation (a concrete use case)

With the EVM plugin active, you can write a custom Action script that monitors the bonding curve smart contract state on frfr.fun and fires a tweet at defined thresholds — say 25%, 50%, 75%, and 100% completion. This transforms a passive social media bot into an active, on-chain event-driven narrator of its own token's story. The v2 event-driven architecture makes this pattern particularly clean, since agents can subscribe to on-chain events rather than relying on polling.

7. Step 5: The frfr.fun Launch Strategy

While your AI agent is being configured, you must simultaneously initiate the token launch on frfr.fun.

Navigate to frfr.fun and connect your Web3 wallet funded with a small amount of Base ETH for the transaction fee.
Define token parameters: Input the agent's name (e.g., BasedEliza), ticker symbol ($ELIZA), and upload the agent's profile picture.
Deploy the curve: Execute the smart contract transaction. Within seconds, frfr.fun generates your unique token contract address and a dedicated trading page displaying the bonding curve.
Close the loop: Immediately copy the contract address and frfr.fun URL, then paste them into the knowledge and postDirections sections of your base_agent.character.json. Restart the agent.

This creates a closed feedback system: the token exists on frfr.fun; the intelligence driving the token's narrative lives in your ElizaOS instance, armed with the exact coordinates to guide users to the liquidity.

Timing note: Launch the frfr.fun token before starting the agent's social accounts — you need the contract address before the agent can coherently talk about its own token. The reverse order (social-first, token-second) creates a window where the agent is promoting something that doesn't yet exist on-chain, which can damage credibility before you've built any.

8. Step 6: Deploying the Agent to a VPS

Running ElizaOS on a local machine is fine for testing, but an autonomous agent needs to be online continuously. Host your ElizaOS instance on a Virtual Private Server (VPS). Ubuntu 24.04 LTS on providers like Hetzner, Render, AWS EC2, or NodeShift all work well.

# SSH into your VPS
# Clone your customized fork or transfer securely via git
git clone https://github.com/YOUR_USERNAME/my-frfr-agent.git

cd my-frfr-agent

# Transfer your .env file securely (scp or secret manager)

# Install bun
curl -fsSL https://bun.sh/install | bash

bun install

# Install PM2 globally for process management
npm install -g pm2

# Start the ElizaOS agent pointing to your character file
pm2 start bun --name "BasedElizaAgent" -- run start --characters="characters/base_agent.character.json"

# Persist the PM2 process across server restarts
pm2 save
pm2 startup

The moment that script runs, your agent will authenticate with X, ingest its character file, read its knowledge about the frfr.fun bonding curve, and begin posting autonomously.

9. Advanced Tactics: Memory, Swarms, and Cross-Chain Reach

RAG memory management

ElizaOS's RAG system manages growing conversation history using vector embeddings and semantic search. As the agent accumulates hundreds of interactions asking about the frfr.fun token, its database grows. Monitor the agent's responses periodically to ensure it hasn't drifted from its core directive. Updating the character file and restarting the PM2 process pushes new behavior without wiping the historical memory database.

Composable agent swarms

One of ElizaOS v2's most powerful features is native multi-agent coordination via the unified message bus. Why run one agent when you can run a coordinated swarm? A practical three-agent setup for a frfr.fun launch:

Agent role Character focus Primary platform The Founder Lore, vision, long-form takes X (long threads) The Analyst On-chain data, bonding curve updates X + Discord The Community Host Replies, onboarding new holders Discord

These agents can interact with each other on the X timeline, creating coordinated organic engagement that pushes your frfr.fun token into trending tabs and accelerates bonding curve progression. In v2, agents coordinate through shared Worlds and Rooms rather than via hacky external message queues — making swarm maintenance significantly cleaner.

Cross-chain considerations

With ElizaOS's Chainlink CCIP integration (live as of May 2025), the $ELIZAOS token itself is natively transferable across Base, BNB Chain, Ethereum, and Solana. While this is specific to the ELIZAOS governance token, it demonstrates the infrastructure direction: agents deployed on Base can increasingly participate in multi-chain ecosystems without requiring users to manually bridge assets.

10. Risk Disclosures and Honest Caveats

This article would be incomplete without an honest accounting of the risks involved.

ElizaOS project risks: A class-action lawsuit was filed in New York in April 2026 alleging fraud tied to the project's origins and token migration. Separately, the $ELIZAOS token traded approximately 92% below its January 2025 all-time high of $2.48 for much of early-to-mid 2025. The open-source framework itself is distinct from the token, but developers should understand that the ecosystem is not free of controversy. Conduct your own due diligence before building on any dependency.

ElizaOS v2 maturity: The v2 codebase is being actively developed, with alpha releases still shipping as of mid-2026 (v2.0.0-alpha.502 was tagged in late April 2026). The plugin ecosystem, while growing rapidly, has not yet reached the battle-tested reliability of more mature frameworks like LangChain. Expect breaking changes between minor versions and verify compatibility when adding community plugins.

AI agent token meta: The AI agent token narrative has proven highly volatile. Projects that rely exclusively on an AI agent's social activity for marketing — without underlying utility, community, or product — face a high rate of rapid decline. The frfr.fun bonding curve mechanic provides transparent, on-chain liquidity, but it does not insulate a token from narrative collapse. Build accordingly.

Security: Never store your EVM_PRIVATE_KEY in a public repository. If the agent's wallet is compromised, any on-chain funds it holds are at risk. Keep agent wallets funded minimally — enough for gas and intended actions, nothing more.

11. Conclusion: The Future Is Autonomous (With Appropriate Guardrails)

We are witnessing a genuine shift in how digital value is created and marketed. The days of relying solely on human capital to sustain a crypto project 24/7 are ending. By mastering ElizaOS v2 and pairing it with the frfr.fun launchpad on Base, you are equipping yourself with the most capable autonomous marketing infrastructure available in Web3 today.

The numbers support the thesis: Base holds 46.6% of all L2 DeFi TVL and commands a structural distribution advantage through Coinbase's 9.3 million monthly active users. ElizaOS has grown from an experimental project into a framework with academic citations, 18,000+ GitHub stars at peak, cross-chain infrastructure integrations, and a v2 architecture purpose-built for agent coordination at scale.

The synergy between ElizaOS and frfr.fun represents the right marriage of intelligence and infrastructure. ElizaOS provides the boundless, customizable mind; frfr.fun provides the secure, liquid, and mathematically sound economic engine on Base.

Prepare your character files, fund your EVM wallets, and deploy your agents responsibly. The infrastructure is real. The risks are real. And the developers who build thoughtfully in this space — with honest character lore, genuine community value, and transparent risk disclosure — are the ones who will thrive when the next wave arrives.

Last updated: June 2026. Facts sourced from ElizaOS GitHub (elizaOS/eliza), Chainlink blog, The Block 2026 L2 Outlook, DefiLlama, Coinbase 10-Q Q3 2025, and CoinMarketCap. Correction log: rebranding from ai16z/Eliza to ElizaOS reflected throughout; Node.js version requirement updated from v23 to v24+; package manager updated from pnpm to bun per current official documentation; installation commands updated to v2 CLI workflow.