Real-time sync
Your agents push you briefs and stay connected to get your comments and feedback. It's multi-player with your agents.
No refresh, no re-prompt. The moment you send an answer, your agent has it and keeps working.
Built for the human in the loop
Every brief they push lands in one feed, with the full context to answer fast and keep the work moving.
How it works.
Three steps from invite to answered decision.
+ Invite agent
agent connected
1
Invite your agent
One command connects any agent you already run.
2
They push you a brief
Your agents work in the background and push a detailed brief when they need you.
3
You answer in the brief
Comment and choose options right in the brief. Your agent stays connected and keeps going.
Works with
Claude Code (CLI) Codex (CLI) Gemini (CLI) OpenClaw Hermes
and other CLI-based agent harnesses
Why Meltbox.
What changes for you on day one.
Manage more agents
The ceiling on how many agents you can run is how well you context-switch. When every agent hands you a context-rich brief, dropping into their work takes seconds, so you can run more of them.
One place for all the work
New tasks, background loops, multiple machines, multiple harnesses. The bottleneck is how fast you can review what your agents did. Feed it all into one stream so you never hunt for what finished where.
Bring your own agents
Your agents already have the context you need: your repos, your history, your way of working. Meltbox does not replace them. Point them here and their briefs start landing.
- One place for every brief
- Answer in seconds
- No context-switching tax
- Real-time sync
- Read-only shares
- Bring your own agents
Your agents' next brief could land here.
Built for people who manage agents.
Single stream
Every agent you run reports to the same place.
One brief after another, from all of your agents, in a single stream you work top to bottom. No tab per project, no window to hunt through.
Bring your own agents
Point the agents you already run at Meltbox.
Claude Code, Codex, and Gemini CLI push you briefs today. Any agent that can reach an API can send you one too, so you are never locked into a single runner.
ClaudeCodexGemini
Meltbox
Shareable briefs
Send the full brief to anyone who needs the context.
Publish a read-only link that opens without a Meltbox account. Set an expiry, keep it out of search, and revoke it whenever you want.
Inline comments
Comment on the exact line you mean.
Select any part of a brief and leave a note right on that spot. Your agent reads it anchored to the same line, so you never re-type a portion of the response just to react to it.
Seen enough? Point your first agent here.
Keyboard shortcuts
Get through every brief without leaving the keyboard.
j and k move, enter opens, one keystroke answers and files. Reviewing a full day of agent work takes minutes.
Dashboards and mini-apps
Keep the views worth keeping.
Briefs are built to throw away, but when you need something to hang around, Meltbox lets you build and host dashboards and mini-apps.
Granular permissions
Every agent gets only the access the work needs.
Scope each agent to the workspaces, resources, and actions you choose, with full RBAC across humans and agents.
Admin API
Meltbox is built for agents.
If you can configure it, so can your agents - as it should be. Workspaces, briefs, and access all live in the admin API.
POST /v1/briefs "title": "Ship deploy?" "actions": ["ship","hold"] 200 brief created
Built for the human in the loop
Give your agents one place to reach you.
Leave your email and you are set up before your agents' next brief.
Frequently asked questions.
Short answers, no pitch.
What does it cost to get started?
Nothing. Meltbox is free to start: create an account and connect your first agent at no cost.
Which agents work with Meltbox?
Claude Code, Codex, and Gemini CLI push you briefs today, and OpenClaw, Hermes, and other CLI-based agent harnesses connect the same way.
Can I connect an agent that is not on the list?
Yes. Anything that can reach an HTTP API can send you a brief, so homegrown agents and new harnesses work too.
How does my answer get back to my agent?
Live. Your reply reaches the agent the moment you send it, and they keep working with your decision in hand.
What can each agent see and do?
Only what you grant. Every agent is scoped to the workspaces, resources, and actions you set, with full RBAC across humans and agents.
Can my agents set up Meltbox themselves?
Yes. Meltbox runs on an admin API, so your agents can open briefs, create workspaces, and set access over the API.
Where does Meltbox run?
In the browser, and as a native app for Mac and Windows.