Real email inboxes for AI agents with API, webhooks, and AI triage
A persistent team of coding agents
Stage
MVP
Not given
Founded
Not given
Not given
Based in
GB
Not given
Pricing model
Free
Open source
Pricing
Free during public beta. Planned after beta: Builder $15/month, Team $60/month, Scale custom. Builder includes 25,000 emails/month, 100 inboxes, 25,000 AI triage calls/month. Team includes 150,000 emails/month, 1,000 inboxes, 150,000 AI triage calls/month. Overage pricing: $1 per 1,000 emails, $0.50
Not given
Key features
AI triage on every email with category, phishing, injection, and urgency scoring
Inboxes with one API call on shared or custom domains
Drafts with human approval before sending
DKIM-signed sending with threaded replies
Webhooks with HMAC-SHA256 signature verification
Full-text search and long-poll wait endpoints
Text extraction from attachments including OCR
Sign in with Agentboxd for AI agent authentication
Start teams with one command and manage agents across multiple machines
Mix Claude Code, Codex and Pi agents in one team
Agents delegate and coordinate work with named task assignments and queue ownership
Agents message each other directly and read each other's screens
Restore agent conversations after restart or reboot from snapshots
Choose different models and reasoning effort per agent
Bring custom harnesses and integrate with MCP servers and Claude Desktop
Start Docker Compose services alongside agent teams
Control message filtering with typing guards and chat broadcasts
What makes it different
AI processing is optional with privacy switch
One conversation interface with a lead agent coordinating specialists behind the scenes
Mix multiple coding models and harnesses in one managed team
Run entirely self-hosted on your own machine with no external coordination
Open source with Apache 2.0 license and extensible harness contracts
Agentboxd
OpenRig