Installation

System Requirements

Operating System

macOS or Linux (Windows not tested)

Python

Python 3.12 (required)

Node.js

Node.js 18+ (for web interface)

Git

Git for version control

Installation Methods

Poetry (Recommended)
pip (Alternative)

Poetry provides the best dependency management and isolation:

Install Poetry

curl -sSL https://install.python-poetry.org | python3 -

Add Poetry to your PATH (if not already):

export PATH="$HOME/.local/bin:$PATH"

Clone Repository

git clone https://github.com/HyperbolicLabs/Hyperbolic-AgentKit.git
cd Hyperbolic-AgentKit

Setup Environment

# Use Python 3.12
poetry env use python3.12

# Install all dependencies
poetry install

# Install Playwright browsers
poetry run playwright install

Using pip requires manual dependency management. Poetry is recommended.

# Create virtual environment
python3.12 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Install Playwright
playwright install

Core Dependencies

The framework includes these major packages:

pyproject.toml

[tool.poetry.dependencies]
python = ">=3.11,<4.0"
coinbase-agentkit = "0.1.6"
coinbase-agentkit-langchain = "^0.1.0"
langchain-openai = "0.3.1"
langgraph = "^0.2.73"
langchain-anthropic = "^0.3.0"
langchain-community = "^0.3.11"
paramiko = "^3.4.0"  # SSH access
tweepy = "^4.15.0"   # Twitter API
browser-use = "^0.1.37"  # Browser automation
anthropic = ">=0.41.0,<1.0.0"  # Writing agent
pypdf = "^4.0.1"     # PDF processing

API Keys Configuration

Required API Keys

Create Environment File

cp .env.example .env

Configure Core Keys

Add these required keys to .env:

.env

# Anthropic - For Claude LLM
ANTHROPIC_API_KEY=sk-ant-api03-xxx

# CDP - For blockchain operations
CDP_API_KEY_NAME=organizations/{org_id}/apiKeys/{key_id}
CDP_API_KEY_PRIVATE=-----BEGIN EC PRIVATE KEY-----\nxxx\n-----END EC PRIVATE KEY-----
CDP_NETWORK_ID=base-mainnet

# Hyperbolic - For GPU compute
HYPERBOLIC_API_KEY=your_hyperbolic_api_key

How to get CDP API keys

Sign up at CDP Portal
Create a new API key
Download the JSON file with your credentials
The name field is your CDP_API_KEY_NAME
The privateKey field is your CDP_API_KEY_PRIVATE
Keep the newlines in the private key (use \n in .env file)

Enable Core Toolkits

.env

# Core toolkits
USE_COINBASE_TOOLS=true
USE_HYPERBOLIC_TOOLS=true
USE_BROWSER_TOOLS=true

Optional API Keys

Twitter/X
Writing Agent
Voice Agent
Podcast Agent
GitHub Integration

Required for social media features:

.env

# Twitter API v2 credentials
TWITTER_ACCESS_TOKEN=your_access_token
TWITTER_API_KEY=your_api_key
TWITTER_API_SECRET=your_api_secret
TWITTER_ACCESS_TOKEN_SECRET=your_access_token_secret
TWITTER_BEARER_TOKEN=your_bearer_token
TWITTER_CLIENT_ID=your_client_id
TWITTER_CLIENT_SECRET=your_client_secret

# Enable Twitter tools
USE_TWITTER_CORE=true
USE_TWITTER_KNOWLEDGE_BASE=true
USE_TWEET_REPLY_TRACKING=true
USE_TWEET_REPOST_TRACKING=true

Get Twitter API Access

Create a developer account and app at Twitter Developer Portal

Required for content generation:

.env

# Tavily for web research
TAVILY_API_KEY=tvly-xxx

# Enable writing agent
USE_WRITING_AGENT=true
USE_WEB_SEARCH=true

The writing agent uses Anthropic API (already configured) and Tavily for web search.

Get Tavily API key at tavily.com

Required for WebSocket voice interface:

.env

# OpenAI for voice capabilities
OPENAI_API_KEY=sk-xxx

Voice agent only works with GPT-4o Realtime models:

gpt-4o-realtime-preview
gpt-4o-mini-realtime-preview

Required for video/podcast processing:

.env

# Google Cloud for Gemini
# Place service account key as eaccservicekey.json in project root
USE_PODCAST_KNOWLEDGE_BASE=true

Create a Google Cloud service account
Download the key as eaccservicekey.json
Place it in the project root directory

.env

GITHUB_TOKEN=ghp_xxx
USE_GITHUB_TOOLS=true

SSH Configuration

SSH access is required for connecting to rented GPU instances.

Generate RSA Key

Only RSA keys are supported (not Ed25519):

ssh-keygen -t rsa -b 4096 -C "[email protected]"

Save to default location: ~/.ssh/id_rsa

Configure in .env

.env

SSH_PRIVATE_KEY_PATH=~/.ssh/id_rsa

Add Public Key to Hyperbolic

Go to app.hyperbolic.xyz
Navigate to Settings
Add your public key (~/.ssh/id_rsa.pub)

Crypto Setup for GPU Compute

To pay for GPU compute with cryptocurrency:

Prepare Wallet

You need an Ethereum wallet with funds on the Base network.Supported tokens:

USDC
USDT
DAI

Connect Wallet

Use the agent to link your wallet:

User: connect my wallet 0xYOUR_WALLET_ADDRESS to Hyperbolic

The agent will use the link_wallet_address action:

# From hyperbolic_agentkit_core/actions/link_wallet_address.py
def link_wallet_address(wallet_address: str) -> str:
    api_key = get_api_key()
    endpoint = "https://api.hyperbolic.xyz/v1/billing/wallet"
    payload = {"wallet_address": wallet_address}
    response = requests.post(endpoint, headers=headers, json=payload)
    return response.json()

Send Funds

Transfer funds to Hyperbolic’s Base address:

To: 0xd3cB24E0Ba20865C530831C85Bd6EbC25f6f3B60
Network: Base
Token: USDC, USDT, or DAI

Funds are available immediately after confirmation.

Start Computing

You can now rent GPUs:

User: Rent an H100 GPU

Character Configuration

Customize your agent’s personality:

Copy Template

cp characters/default.json characters/my-agent.json

Edit Configuration

characters/my-agent.json

{
  "name": "MyAgent",
  "bio": [
    "AI assistant specialized in GPU compute",
    "Expert in blockchain and DeFi"
  ],
  "lore": [
    "Built to help developers access compute resources"
  ],
  "adjectives": [
    "helpful",
    "technical",
    "efficient"
  ],
  "style": {
    "all": [
      "Be concise and technical",
      "Use clear examples"
    ]
  },
  "topics": [
    "AI/ML infrastructure",
    "GPU computing",
    "Blockchain"
  ]
}

Use Custom Character

.env

CHARACTER_FILE=characters/my-agent.json

Tool Configuration

Enable or disable specific tools:

.env

# Core Toolkits
USE_COINBASE_TOOLS=true
USE_HYPERBOLIC_TOOLS=true
USE_BROWSER_TOOLS=true

# Twitter Tools
USE_TWITTER_CORE=true
USE_TWEET_REPLY_TRACKING=true
USE_TWEET_REPOST_TRACKING=true
USE_TWEET_DELETE=true
USE_USER_ID_LOOKUP=true
USE_USER_TWEETS_LOOKUP=true
USE_RETWEET=true

# Additional Tools
USE_WRITING_AGENT=true
USE_WEB_SEARCH=true
USE_REQUEST_TOOLS=true
USE_GITHUB_TOOLS=true

# Knowledge Bases
USE_TWITTER_KNOWLEDGE_BASE=false
USE_PODCAST_KNOWLEDGE_BASE=true

# Security
ALLOW_DANGEROUS_REQUEST=true  # Required for RequestsToolkit

Running the Application

Terminal Chat
Gradio Web UI
Voice Agent

Interactive command-line interface:

poetry run python chatbot.py

Choose mode:

1 - Interactive chat mode
2 - Character Twitter automation

Graphical web interface:

poetry run python gradio_ui.py

Access at http://localhost:7860

WebSocket-powered voice interface:

PYTHONPATH=$PWD/server/src poetry run python server/src/server/app.py

Access at http://localhost:3000Available voices:

alloy
ash
ballad
coral
echo
sage
shimmer
verse

Verification

Test your installation:

from hyperbolic_langchain import HyperbolicToolkit, HyperbolicAgentkitWrapper

hyperbolic = HyperbolicAgentkitWrapper()
toolkit = HyperbolicToolkit.from_hyperbolic_agentkit_wrapper(hyperbolic)
tools = toolkit.get_tools()

print(f"Loaded {len(tools)} Hyperbolic tools:")
for tool in tools:
    print(f"  - {tool.name}")

Troubleshooting

Dependencies Install Failed

Clean Poetry cache and reinstall:

poetry env remove python3.12
poetry env use python3.12
poetry install --no-cache

Browser Automation Issues

Reinstall Playwright with system dependencies:

poetry run playwright install --force
poetry run playwright install-deps

Chrome Profile Issues

For browser automation:

Ensure Google Chrome is installed
Remove all pinned tabs from default profile
Disable profile selector on startup
Set Chrome as default browser

SSH Connection Failed

Verify SSH key configuration:

# Test key exists
ls -la ~/.ssh/id_rsa

# Check key type (should be RSA)
head -1 ~/.ssh/id_rsa
# Should show: -----BEGIN RSA PRIVATE KEY-----

# Ensure public key is in Hyperbolic settings
cat ~/.ssh/id_rsa.pub

API Rate Limits

If you hit rate limits:

Twitter: Client has wait_on_rate_limit=True built-in
Anthropic: Use ANTHROPIC_MAX_RETRIES env var
Hyperbolic: Contact support for limit increases

Next Steps

Build Your First Agent

Follow the quickstart guide

Explore Tools

Learn about available tools and capabilities

Customize Behavior

Configure agent personality and style

Agent Types

Explore different agent interfaces

Get Started

Core Concepts

Guides

Agent Types

System Requirements

Operating System

Python

Node.js

Git

Installation Methods

Core Dependencies

API Keys Configuration

Required API Keys

Optional API Keys

Get Twitter API Access

SSH Configuration

Crypto Setup for GPU Compute

Character Configuration

Tool Configuration

Running the Application

Verification

Troubleshooting

Next Steps

Build Your First Agent

Explore Tools

Customize Behavior

Agent Types

Build docs developers (and LLMs) love

Get Started

Core Concepts

Guides

Agent Types

​System Requirements

Operating System

Python

Node.js

Git

​Installation Methods

​Core Dependencies

​API Keys Configuration

​Required API Keys

​Optional API Keys

Get Twitter API Access

​SSH Configuration

​Crypto Setup for GPU Compute

​Character Configuration

​Tool Configuration

​Running the Application

​Verification

​Troubleshooting

​Next Steps

Build Your First Agent

Explore Tools

Customize Behavior

Agent Types

Build docs developers (and LLMs) love

System Requirements

Installation Methods

Core Dependencies

API Keys Configuration

Required API Keys

Optional API Keys

SSH Configuration

Crypto Setup for GPU Compute

Character Configuration

Tool Configuration

Running the Application

Verification

Troubleshooting

Next Steps