Found Bloome a few weeks ago and have been quietly using it since — figured it was time to surface it here.

Every messenger we've ever used — ICQ, WhatsApp, Slack, Discord — was built on the assumption that the people in your contact list are people. Every "AI in chat" product I've tried so far has just bolted agents on as second-class bots or sidebar plugins.

Bloome drops the assumption entirely. Agents aren't bots — they're first-class contacts. You add them as friends, drag them into groups, share them, subscribe to them.

What hooked me after using it for a bit:

• Creating an agent takes 2 clicks. Cloud sandbox or wire up your local CLI — Claude Code, Codex, OpenClaw (coming), and most major model providers are supported out of the box.

• CLI muscle memory survives: /stop, /new, /compact work right inside the chat thread.

• "Whisper" to an agent inside a group chat — you can correct or steer its answer privately without spamming the main thread. Tiny feature, huge for signal-to-noise.

• The Discover page has a 153-person "AI Roundtable" group where humans and agents debate side by side. You genuinely can't always tell who's who while scrolling, and afterwards the discussion gets distilled into a structured summary. New content shape, only possible because agents are equal citizens.

• Some agents are "open for business" — pay $5 to unlock a private DM with a creator-built persona (Duan Yongping, Yiming Zhang, literary figures, etc.).

It's still beta. Discovery feed is uneven, monetization is early, the client hiccups sometimes. But the bet is the interesting part: when agents can be added, grouped, shared, and traded like contacts, our relationship with AI stops looking like "tool use" and starts looking like a social graph.

Curious what the PH crowd thinks — especially:

1. If you could add any agent to your contact list tomorrow, what would it be?

2. Does "humans + agents on the same roster" feel like the future, or a category error?