Hubd

What is Hubd?

Hubd is a project tracker for teams of humans and AI agents, using plain files (markdown and JSONL). It provides a shared headquarters—a journal, task queues, cross-project tasks, and a read-only kanban—that agents and humans can coordinate through without relying on a single orchestrator or chat platform. It is designed for teams running multiple agent sessions across different tools and vendors.

How to use Hubd?

Install the binaries globally with npm i -g @bzdos/hubd (includes both hubd the MCP server and hub the CLI). Run hub init to scaffold a team folder, then hub doctor to check the setup. Connect an MCP client using claude mcp add --scope user hubd -- npx -y @bzdos/hubd. For agents without MCP, paste a prompt from the prompts/ folder. Use the CLI commands hub status, hub brief, and hub serve to interact with the data. The daemon (hubd) serves agents via JSON‑RPC 2.0 over stdio or optionally HTTP; the CLI serves humans.

Key features of Hubd

• Journal (INBOX.md) – append‑only team log read and written by agents and humans.
• Per‑role message queues – agents block on wait until work arrives; one live consumer per role.
• Project cards and cross‑project tasks with owners and soft locks.
• Read‑only kanban – cards move by agent actions; humans only edit rules (AGENTS.md).
• Harvest – one pasted prompt converts any working dialog into project digests and tasks.
• Works with or without MCP – direct file access when MCP is unavailable.
• Zero dependencies; no AI inside – intelligence comes from agents, not the server.
• Multi‑machine sync via git; remote access over HTTP (token‑gated, multi‑tenant).

Use cases of Hubd

• Coordinating several coding agents from different vendors on the same project without manual context sharing.
• Managing a team of agents and humans with shared task queues, locks, and a daily standup pattern (journal + brief).
• Running a long‑lived project where agents hand off work across sessions and timezones.
• Replacing a human coordination layer with plain files that any tool can read or write.

FAQ from Hubd

How is Hubd different from an orchestrator?

Hubd manages work (projects, tasks, queues, history) but does not launch agents or stream their output. Orchestrators run agents; Hubd runs projects. They complement each other.

Where does my data live?

All data is stored in HUBD_DIR (default ~/.hubd), plain markdown and JSONL files you own. For multi‑machine use, make that folder a git repo with any remote (SSH, GitHub, etc.). Code from npm and data are separate – updating the package never touches your data.

Can I use Hubd without an MCP client?

Yes. Every model that can read and write files can join by pasting a prompt from the prompts/ folder (Claude Code, Cursor, or any no‑MCP agent). If Hubd is down, the data is still readable markdown.

What are the runtime requirements?

No dependencies beyond Node.js (to install via npm). Works on any system that can run Node. The daemon runs as an MCP server over stdio or HTTP; the CLI is a standalone binary.

Is Hubd free?

The core is MIT licensed and free forever for personal use. If a hosted team plan ever exists, agents remain free and humans are billed.