Skip to main content

​
General

Hive is an open-source agent framework that lets you build autonomous, reliable, self-improving AI agents without hardcoding workflows. You define your goal through conversation with the Hive coding agent (queen), and the framework generates a node graph with dynamically created connection code. When things break, the framework captures failure data, evolves the agent, and redeploys.
Hive is designed for developers and teams who want to build production-grade AI agents without manually wiring complex workflows.Hive is a good fit if you:
  • Want AI agents that execute real business processes, not demos
  • Need fast or high volume agent execution
  • Need self-healing and adaptive agents that improve over time
  • Require human-in-the-loop control, observability, and cost limits
  • Plan to run agents in production environments
Hive generates your entire agent system from natural language goals using a coding agent—you don’t hardcode workflows or manually define graphs. When agents fail, the framework automatically captures failure data, evolves the agent graph, and redeploys. This self-improving loop is unique to Aden.
Traditional FrameworksHive
Hardcode agent workflowsDescribe goals in natural language
Manual graph definitionAuto-generated agent graphs
Reactive error handlingOutcome-evaluation and adaptiveness
Static tool configurationsDynamic SDK-wrapped nodes
Separate monitoring setupBuilt-in real-time observability
DIY budget managementIntegrated cost controls & degradation
Yes, Hive is fully open-source under the Apache License 2.0. We actively encourage community contributions and collaboration.

​
LLM Providers & Models

Hive supports 100+ LLM providers through LiteLLM integration, including:
  • OpenAI (GPT-4, GPT-4o)
  • Anthropic (Claude models)
  • Google Gemini
  • DeepSeek
  • Mistral
  • Groq
  • And many more
Simply set the appropriate API key environment variable and specify the model name. We recommend using Claude, GLM, and Gemini as they have the best performance.
Yes! Hive supports local models through LiteLLM. Simply use the model name format:
model="ollama/llama3"
model="ollama/mistral"
Ensure Ollama is running locally before using local models.
Yes! v0.5.1 added first-class support for Claude Code subscriptions. Set use_claude_code_subscription: true in ~/.hive/configuration.json and the framework will read OAuth tokens from ~/.claude/.credentials.json with automatic refresh.
Yes! You can configure api_base in your ~/.hive/configuration.json to route traffic through any OpenAI-compatible API, including:
  • Azure OpenAI
  • vLLM
  • Ollama
  • Custom inference servers

​
Features & Capabilities

Yes. Hive is explicitly designed for production environments with features like:
  • Automatic failure recovery
  • Real-time observability
  • Cost controls
  • Horizontal scaling support
The framework handles both simple automations and complex multi-agent workflows.
Yes, Hive fully supports human-in-the-loop (HITL) workflows through intervention nodes that pause execution for human input. These include:
  • Configurable timeouts
  • Escalation policies
  • Multiple question types (free text, structured, selection, approval, multi-field)
  • State persistence across pause/resume
This allows seamless collaboration between human experts and AI agents.
Yes. Aden’s SDK-wrapped nodes provide built-in tool access, and the framework supports flexible tool ecosystems. Agents can integrate with:
  • External APIs
  • Databases
  • File systems
  • Web services
  • MCP (Model Context Protocol) servers
Hive includes 100+ pre-built tools covering file operations, web scraping, data processing, communication (email, Slack, Discord), CRM integrations, and more.
Hive provides granular budget controls including:
  • Spending limits
  • Throttles
  • Automatic model degradation policies
You can set budgets at the team, agent, or workflow level, with real-time cost tracking and alerts. Token tracking for reasoning and cache tokens is on the roadmap.
Yes! v0.5.1 introduced multi-graph agent sessions, allowing you to load, manage, and switch between multiple agent graphs within a single session. The Hive Coder uses multi-graph internally to coordinate between worker agents and supervisory agents.
Yes, this is a core feature of Hive. When agents fail, the framework:
  1. Captures failure data
  2. Analyzes the failure against defined goals
  3. Evolves the agent graph through the Hive Coder
  4. Redeploys the improved agent
This creates a continuous improvement loop without manual intervention.

​
Development & Deployment

The Hive framework is built in Python 3.11+. A JavaScript/TypeScript SDK is on the roadmap.
Yes, Hive is designed for self-hosting. The framework includes:
  • Docker support with Python 3.11-slim base
  • Playwright Chromium for browser automation
  • Health check endpoints
  • Volume mounts for persistence
  • Webhook server for external triggers
See the self-hosting guide for complete setup instructions.
Yes! Hive provides official Docker support with:
  • Python 3.11-slim base image
  • Playwright Chromium installation
  • Non-root user for security
  • Health check endpoint
  • Port 4001 exposed for MCP server
  • Volume mount for workspace persistence
Yes! Hive supports multiple entry points:
  • Manual - Direct user invocation
  • Webhook - HTTP POST triggers
  • Timer - Cron expressions or interval_minutes
  • Event - Custom event triggers
  • API - Programmatic invocation
Each entry point supports configurable isolation levels (isolated, shared, synchronized).
Hive includes a comprehensive testing framework:
# Run lint and format checks
make check

# Run core framework tests
make test

# Run tests for a specific agent
PYTHONPATH=exports uv run python -m agent_name test
The framework supports:
  • Level 0: Output key validation
  • Level 1: Literal checks (output_contains, output_equals)
  • Level 2: Conversation-aware LLM judgment
  • Custom test cases with success criteria

​
Getting Started

# Clone the repository
git clone https://github.com/aden-hive/hive.git
cd hive

# Run quickstart setup
./quickstart.sh
This sets up:
  • Core framework and MCP tools
  • Encrypted credential store
  • LLM provider configuration
  • All required Python dependencies
Hive uses a uv workspace layout and is NOT installed with pip install. Running pip install -e . will create a placeholder package and Hive will not function correctly.
The easiest way is to use the Hive Coder:
# Launch the Coder directly
hive code

# Or from TUI
hive tui
# Press Ctrl+E to escalate to Coder
Describe what you want your agent to do, and the Hive Coder will:
  1. Generate the goal definition with success criteria
  2. Create the node graph and connection code
  3. Wire in the appropriate tools
  4. Generate test cases
  5. Export a runnable agent package
Hive includes 100+ pre-built tools:
  • File Operations (36+ tools) - read, write, edit, search, diff/patch
  • Web Tools - Search, scraper, Exa, News, SerpAPI
  • Data Tools - CSV, Excel, PDF, Vision, Time
  • Communication - Email, Gmail, Slack, Discord, Telegram, Google Docs
  • CRM/API - HubSpot, GitHub, Apollo, BigQuery, Razorpay, Cal.com
  • Security/Scanning - DNS, SSL/TLS, Port Scanner, Subdomain Enumerator
All tools support MCP (Model Context Protocol) for easy integration.

​
Contributing & Support

Contributions are welcome! Follow these steps:
  1. Find or create an issue and get assigned
  2. Fork the repository
  3. Create your feature branch
  4. Make your changes and add tests
  5. Ensure make check and make test pass
  6. Submit a Pull Request
See the contributing guide for detailed guidelines.
We’re especially looking for help building tools, integrations, and example agents. Check issue #2805 for opportunities.
Please do NOT report security vulnerabilities through public GitHub issues.Instead, email: [email protected] with:
  1. A description of the vulnerability
  2. Steps to reproduce the issue
  3. Potential impact
  4. Any possible mitigations you’ve identified
We will acknowledge receipt within 48 hours and aim to resolve critical vulnerabilities within 7 days.See our security policy for details.

​
Still Have Questions?

Join our Discord community or open an issue on GitHub.

Build docs developers (and LLMs) love

Get started for freeTalk to us