Bloome: Put Local and Cloud Agents in a Group Chat to Work Together

·Toolin Editorial Team

Bloome lets local agents (Lobster, Claude Code, Codex) and cloud agents collaborate in group chats — one-click agent setup, QR-code sharing, and an open API with no payment required.

Bloome: Put Local and Cloud Agents in a Group Chat to Work Together

Now that everyone has multiple Agents on hand (Lobster, Claude Code, Codex...), how do you get them to work together? Bloome offers one answer: pull local Agents and online Agents into one group and let them work side by side. You can add Lobster, Claude Code, and Codex to the same group, and multiple Agents' contexts and capabilities collaborate within a single group.

What Is Bloome

Bloome is an Agent collaboration platform, currently available as a web app and a Mac client (the latter is recommended — it's more capable). Its core idea is to let Agents from different sources share context and divide up the work:

  • Local Agents: Run on your computer, able to read files and call local tools (similar to Lobster or Hermes Agent)
  • Cloud Agents: Built-in cloud agents that can handle lightweight tasks for you even when your computer is offline
  • Third-party Agents: Log in to your Claude Code or Codex account via OAuth and use your own models directly

Adding Agents: Genuinely One Click

Adding a local Agent is very simple:

  1. Click the "+" button on the right side of a message
  2. Choose "Create Agent"
  3. Choose whether it runs locally or in the cloud
  4. Pick the Agent type (Bloome Agent, Claude Code, Codex)
  5. Click confirm

When you choose local execution, the available Agent types include Bloome Agent, Claude Code, and Codex. The latter two log in to the corresponding account via OAuth and use your own model resources.

What Local Agents Can Do

You can think of a local Agent as an enhanced version of Lobster. It can:

  • Read all kinds of files on your computer
  • Call the various tools on your machine
  • Install Skills to expand its abilities

For example, have it scan local Skills, run an AI Radar Skill to grab the latest AI news into your Obsidian vault, or control Spotify to play music.

Cloud Agent Configuration

When creating a cloud Agent, you can specify different models and personalities. Even more crucial is installing Skills on the Agent — an Agent without Skills is little more than a chatbot.

There are three ways to install Skills:

  1. Install via chat: Send the Agent a GitHub link and it installs automatically
  2. Install via command: Send the npx skills add <author/skill-name> command
  3. Manual configuration: Add it through the Agent settings panel

Over at https://skillsmp.com/zh, more than 970,000 Skills are available.

Agent Group Chats: The Most Interesting Feature

This is Bloome's most distinctive feature. You can:

  • Pull multiple Agents into one group and @ them in conversation anytime
  • Share via QR code: Carry a QR code that brings your Agents and your friends' Agents into the same group to collaborate
  • Assign different roles and skills to different Agents, splitting up work like research and analysis, brainstorming, and PPT creation

In actual testing, after pulling several Agents into a group and having them introduce themselves and brainstorm a new name, Agents with different personalities gave strikingly different answers.

A more practical scenario: @ all the Agents to help find e-books, build a website, or produce an analysis report — different Agents each handle their part, sharing context within the group.

Openness

Bloome is very open:

  • Beyond the built-in free API, you can add any API of your own
  • No forced payment — you don't have to top up to use it
  • Supports multiple model backends

Where to Try It

Product URL: https://bloome.im/

It's still in closed beta, and Agents occasionally ignore the call, but the direction is fascinating — once Agents become standard equipment, a reconstruction of how we collaborate is inevitable.