agent-hub

Free

Local-first multitask board for routing, sequencing, and observing repo-local coding agents across projects.

FreeFree tier

Type

Open Source

Visit Website GitHub (38 stars)

About agent-hub

Agent-hub is a local-first, open-source multitask board designed for coordinating coding agents across projects. It provides a lightweight SQLite-backed queueing system for routing tasks to repo-local agents (like Claude Code, Codex, or Kimi Code) through project-backed commands. The tool models task dependencies, retries, blocking, and human handoff, and offers a small HTTP surface with a browser dashboard for real-time visibility. Built entirely on the Python standard library, it keeps dependencies minimal and remains suitable for single-operator workflows in local environments. It is not a hosted platform or a replacement for coding agents, but a control plane for sequencing and observing multiple tasks.

Key Features

SQLite-based task queue for local-first persistence

Dependency-aware dispatch with retries and blocking

Human inbox for manual handoff and review

Task templates and saved queries for reusable workflows

Pipelines for sequencing multi-step agent tasks

Thin browser dashboard for real-time multitask visibility

Small HTTP surface for programmatic control

Python standard library only, minimal external dependencies

Pros & Cons

Pros

Local-first: no cloud dependency, privacy-friendly
Lightweight and portable: uses only Python standard library
Free and open source (MIT license)
Explicit dependency handling and retry logic reduces manual coordination
Human inbox enables safe handoff between agents and human operators
Thin dashboard provides clear multitask visibility without complexity
Designed to work with existing coding agents, not replace them

Cons

Not production-ready or stable for long-term API commitments
Single-operator focus limits multi-user collaboration
Requires manual setup and configuration of local agents
No hosted version; runs only in local environments
Limited to queueing and routing, not a full workflow engine

Best For

Supervising multiple coding agents working on different local repos simultaneouslyQueueing and routing tasks for Claude Code, Codex, Kimi Code, or Qwen CodeOrchestrating multi-step code implementation pipelines with dependency managementHuman-in-the-loop review and handoff for agent-generated code changesLightweight task board for personal or team-side coding workflows

FAQ

What is agent-hub?

Agent-hub is a local-first multitask board that queues, routes, and observes coding agent tasks across projects. It uses SQLite for persistence and provides a thin browser dashboard for visibility.

How does agent-hub differ from coding agents like Claude Code?

Agent-hub is not a coding agent itself. It acts as a control plane that coordinates tasks from a human operator, routing them to repo-local agents and handling dependencies, retries, and handoff.

Is agent-hub production-ready?

No. The current release (v0.1.0) is a public-safe OSS MVP suitable for demos and local exploration, not a mature hosted platform or stable API surface.