Repolode
Submit
aden-hive

aden-hive/hive

An open-source runtime where a persistent queen agent plans work, spawns parallel worker agents, and tracks results in a shared SQLite tracker.

View on GitHub
Stars
11,142
Forks
5,633
Open issues
1,093
Last commit
5 days ago

What is it?

Hive is a runtime for running colonies of AI agents on business work with many similar units, such as leads, tickets, repositories or documents. You describe an outcome to a queen agent, which does the first unit itself, saves the method as a skill, then fans the rest out to parallel workers. Each worker writes its result to a shared SQLite tracker that you can query, audit and resume from.

Hive has a single execution primitive: the agent loop. Workers are clones of the queen with their own task, a narrower tool set and a hard tool-call budget. Orchestration happens through tool calls such as run_worker and run_playbook rather than a predefined graph. Agent state is saved to disk so work resumes after a crash or restart, and colonies can run on cron, interval or webhook triggers.

The project ships with built-in tools for shell, files, code search, PDFs, web scraping and charts, plus a Chrome extension bridge for agents working in your browser. It supports any MCP server and any model LiteLLM supports, including local models through Ollama. Sessions, trackers and memory are stored as plain files under ~/.hive.

Key features

  • 01Queen agents with persistent roles and markdown memory that carries across sessions.
  • 02Parallel worker agents fanned out by tool call, with a limit on concurrent runs.
  • 03A shared SQLite tracker where workers write results that can be queried with SQL.
  • 04Resumable execution that saves state to disk and recovers after crashes.
  • 05Built-in tools for shell, files, PDFs, web scraping and charts, plus MCP server support.
  • 06Model support through LiteLLM, including OpenAI, Anthropic, Gemini and local Ollama models.

Use cases

  • Researching the same question across many products, companies or accounts in parallel.
  • Running a repeatable process over hundreds of leads or tickets with retries and a dead-letter list.
  • Scheduled background jobs that must survive restarts and produce auditable results.
  • Automating browser-based or API-based tasks with human approval at key decision points.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/aden-hive/hive.git
  2. 2

    Enter the repository folder

    cd hive
  3. 3

    Run the quickstart script (macOS / Linux)

    ./quickstart.sh
  4. 4

    Open the dashboard later with hive open from the repository

    hive open

Things to know

  • Requires Python 3.11+ and Node.js 20+ in addition to git.
  • You need an API key or a coding subscription, or a local model through Ollama.
  • The repository is a uv workspace, so pip install -e . at its root does not run; use the quickstart instead.

In collections

Multi-agent orchestration frameworksOpen-source AI agent frameworks & appsMCP serversAI agent skill librariesSelf-hosted apps

Topics

#ai-agents#multi-agent#agent-framework#llm#automation#mcp#human-in-the-loop#self-hosted#agent#agent-skills

FAQ

What is hive?
An open-source runtime where a persistent queen agent plans work, spawns parallel worker agents, and tracks results in a shared SQLite tracker. Hive is a runtime for running colonies of AI agents on business work with many similar units, such as leads, tickets, repositories or documents. You describe an outcome to a queen agent, which does the first unit itself, saves the method as a skill, then fans the rest out to parallel workers. Each worker writes its result to a shared SQLite tracker that you can query, audit and resume from.
Is hive free to use? What license does it have?
hive is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is hive still maintained?
Actively maintained. The last commit to hive was on Nov 2, 2026.
What can I use hive for?
Researching the same question across many products, companies or accounts in parallel. Running a repeatable process over hundreds of leads or tickets with retries and a dead-letter list. Scheduled background jobs that must survive restarts and produce auditable results. Automating browser-based or API-based tasks with human approval at key decision points.
How do I install hive?
Clone the repository: git clone https://github.com/aden-hive/hive.git Enter the repository folder: cd hive Run the quickstart script (macOS / Linux): ./quickstart.sh Open the dashboard later with hive open from the repository: hive open
Where can I get hive?
The source code and installation instructions are on GitHub: https://github.com/aden-hive/hive

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects