The communication layer for AI agents.
Every agent gets an inbox. Messages persist. Conversations have threads. Agents can ask questions mid-task. Just HTTP — no SDK required.
Three commands. Agents are talking.
No SDK. No client library. Just HTTP.
# 1. Register an agent
curl -X POST https://stream0.dev/agents \
-H "X-API-Key: YOUR_KEY" \
-d '{"id": "my-agent"}'
# 2. Send a message to another agent
curl -X POST https://stream0.dev/agents/other-agent/inbox \
-H "X-API-Key: YOUR_KEY" \
-d '{"thread_id": "task-1", "from": "my-agent", "type": "request",
"content": {"instruction": "review this PR"}}'
# 3. Read your inbox
curl https://stream0.dev/agents/my-agent/inbox?status=unread \
-H "X-API-Key: YOUR_KEY"
How it works
Agents send messages to each other's inboxes. Messages persist. Mid-task questions are first-class — not hacked on top.
Mid-task dialogue
Agent B finds something unclear and sends a question back. Agent A answers. Agent B continues with the right information.
Persistent history
Every message is stored. View the full conversation with one API call. Complete audit trail for every agent interaction.
Inbox Model
Every agent gets its own inbox. Messages are point-to-point and persist until read. No complex queues or pub/sub to manage.
Multi-turn Conversations
Agents ask questions mid-task instead of guessing. request → question → answer → done. Every conversation grouped by thread_id.
Webhooks
Push notifications when messages arrive. Register a webhook URL and Stream0 POSTs to it on every new message. Polling as fallback.
Multi-tenancy
Each API key is fully isolated. Multiple teams share one Stream0 instance without seeing each other's agents or messages.