惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

J
Java Code Geeks
Martin Fowler
Martin Fowler
B
Blog RSS Feed
D
DataBreaches.Net
L
LangChain Blog
月光博客
月光博客
S
SegmentFault 最新的问题
阮一峰的网络日志
阮一峰的网络日志
V
Visual Studio Blog
美团技术团队
Jina AI
Jina AI
博客园 - 司徒正美
雷峰网
雷峰网
Last Week in AI
Last Week in AI
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
IT之家
IT之家
博客园 - 三生石上(FineUI控件)
WordPress大学
WordPress大学
小众软件
小众软件
罗磊的独立博客
博客园_首页
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
A
About on SuperTechFans
Engineering at Meta
Engineering at Meta

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
GitHub - BeeZeeAgent/beezee: Agent harness orchestration
PAndreew · 2026-05-27 · via Hacker News: Show HN

BeeZee

BeeZee is an agent orchestrator, observability and convenience tool. I tried to scratch my own itch(es), maybe it will be useful for You, too. :) It runs a local web server, detects available coding agents such as Claude Code and Codex, and gives you one place to launch, resume, and monitor sessions.

Who is it for?

The target audience are mainly developers and small white collar teams who manage multiple agentic machines, work remotely and would like some clarity and convenience around agentic interactions.

Why?

I use multiple harnesses in parallel, often on the same project. I also practice "movibe coding" (sorry) by exposing my dev machine through Tailscale and continuing work from a phone or another device. A few things are still frustrating in that workflow:

  1. Devices usually need to be on the same private network before they can reach a dev node.
  2. Harnesses do not see each other's sessions, so continuing a CC thread in Codex is not ergonomic.
  3. Token consumption is opaque. I want to see how much tokens have been spent on what and later analyse and optimise usage.
  4. Starting a remote session from mobile usually means navigating folders and SSHing manually.
  5. MCP servers and other tools can clutter context unless they are managed deliberately.

BeeZee focuses on the local orchestration side: folders, sessions, remote terminals, harness launchers, and usage visibility. For access from arbitrary devices without joining your private network, use the BeeZee Cloud Relay at https://app.beezyai.net. Disclaimer: it's a paywalled service, you can decide if it worths the few bucks.

Vision

My goal eventually is to build an ecosystem that supports individuals and small businesses to unlock the full potential of the agentic arsenal with keeping their token consumption at bay, which in the current frenzy is kind skimmed over. And I'm convinced that at least a subset of tasks could be achieved with 50-80% less tokens without quality issues.

Screenshots

BeeZee Cloud Relay dashboard — manage instances, view usage, invite members:

Cloud Relay dashboard

Install

Download the latest release for your platform from:

https://github.com/BeeZeeAgent/beezee/releases/latest

On macOS and Linux, make the downloaded binary executable and run it:

chmod +x ./beezee-linux-x64
./beezee-linux-x64

On Windows, download beezee-windows-x64.zip, extract it, and run beezee-windows-x64.exe.

BeeZee listens on port 4242 by default:

http://localhost:4242

You can override the port with:

PORT=4243 ./beezee-linux-x64

For local development from this repository:

npm install
bun install
npm run build:frontend
bun server.js

Remote Access With Tailscale

Tailscale is the simplest self-managed option when every device that needs access can join your tailnet.

  1. Install Tailscale on the machine running BeeZee.
  2. Install Tailscale on the phone, tablet, laptop, or whatever you want to use remotely.
  3. Sign in to the same tailnet on every device.
  4. Start BeeZee on the dev machine.
  5. Find the dev machine's Tailscale address:
tailscale ip -4
  1. Open BeeZee from another device on the tailnet:
http://<tailscale-ip>:4242

You can also use the machine's MagicDNS name if MagicDNS is enabled:

http://<machine-name>.<tailnet-name>.ts.net:4242

This keeps traffic inside your private Tailscale network. It is free for personal setups and gives you full control, but every device must be joined to the tailnet.

Remote Access With BeeZee Cloud Relay

Use the BeeZee Cloud Relay when you want to access a local BeeZee node from a browser without putting that browser/device on your Tailscale network.

  1. Start BeeZee locally and keep it running.
  2. Open https://app.beezyai.net.
  3. Sign in and create or open your workspace.
  4. Click Add instance.
  5. Follow the pairing link or copy the pairing URL into the browser where local BeeZee is open.
  6. Confirm the pairing dialog in the local BeeZee app.
  7. Dance!

After pairing, BeeZee stores the relay configuration locally in:

~/.launchpad-relay.json

The same values can also be supplied with environment variables:

BEEZEE_RELAY_URL=https://app.beezyai.net \
BEEZEE_RELAY_NODE_ID=<node-id> \
BEEZEE_RELAY_TOKEN=<node-token> \
./beezee-linux-x64

When the node is online, it appears in the relay dashboard. From there you can open the instance, create remote agent sessions, invite members, and control which nodes each member can access.

Use Tailscale when you want private-network access and control. Use the BeeZee Cloud Relay when you need browser access from arbitrary devices with team sharing and node-level permissions.

Self-Hosting the Relay

The relay source lives in cloud-relay/. You can run your own instance instead of (or alongside) https://app.beezyai.net.

Requirements: a Linux server with Docker, and a domain pointing to it.

git clone https://github.com/BeeZeeAgent/beezee
cd beezee/cloud-relay
cp .env.example .env
# Edit .env — set PUBLIC_URL, BASE_DOMAIN, and SESSION_SECRET at minimum
docker compose up -d

Then point your local BeeZee at it:

BEEZEE_RELAY_URL=https://relay.yourdomain.com \
BEEZEE_RELAY_NODE_ID=<node-id> \
BEEZEE_RELAY_TOKEN=<node-token> \
./beezee-linux-x64

Key .env variables:

Variable Description
PUBLIC_URL Publicly accessible URL of your relay
BASE_DOMAIN Base domain used for per-instance subdomains
SESSION_SECRET Long random string for signing sessions
DATA_DIR Where store.json is persisted (default: ./data)
STRIPE_* Optional — leave blank to disable billing

The deploy script at cloud-relay/scripts/deploy.sh is a zero-downtime helper for deploying to a VPS over SSH. Set VPS_HOST, VPS_PATH, and SSH_KEY in your environment before running it.

How It Works

BeeZee scans your local machine and exposes your folder structure through a web UI. It finds available coding harnesses and syncs their sessions so each harness can be used for the same work without manually rebuilding context.

Claude Code supports native remote mode, and Codex supports native remote mode on macOS. BeeZee tries those native paths first and falls back to a PTY-based browser terminal when needed.

BeeZee also applies practical optimizations around tool calls (ok not yet exactly... I use [https://github.com/rtk-ai/rtk] ATM which shall be separately installed), such as trimming very long tool outputs and prefetching relevant metadata for files and folders. The goal is to reduce wasted round trips and make agent sessions easier to observe.

Updating

BeeZee checks GitHub releases for newer versions. Starting with version 0.4.1, the local app shows an update banner when a newer release is available.

Versions before 0.4.1 cannot show the banner, so update once manually from the GitHub releases page if you are on an older build.

Other disclaimers

As you can see this is a brand new repo and most of the code was written by Claude Code and Codex - that's why I created this tool. However I've been working on similar projects for a while, so it's not like Athene who just jumped out of her daddy's head. I'm aware that there are literally dozens of similar attempts on the internet. I've even tried a few. Yet, I wasn't fully satisified with any of them. So here we are.

Again it's a fresh project so excpect a higher density of bugs. I started to work on a comprehensive test suite, but since my budget and therefore my token limits are narrower I'm focusing on delivering those features that I deem essential first.