Features
Docker-compatible container runtime with sub-200ms boot
Firecracker microVM orchestration with <125ms spin-up
AI agent sandbox with full isolation (network, disk, I/O, syscall)
Real-time observability of syscalls, network requests, file writes
Native Apple Silicon support (M1–M4, macOS 15+)
Open source (MIT/Apache 2.0/AGPL 3.0) with full source on GitHub
SwiftUI desktop app with Rust runtime backend
CLI tool (abctl) for scripting and automation
Read-only rootfs for guest filesystems
Bridge network isolation with controlled outbound access
Ephemeral, scoped filesystems for sandboxes
No host access from guest environments
OCI compatible (Docker images)
DevContainer support (coming soon)
Discord community with fast responses
Trace visibility for agent steps (prompts, retrievals, tool calls, outputs)
LLM-as-judge evaluation for relevance, toxicity, quality scoring
Dataset creation from traces for reproducible testing
Experiment management and regression benchmarking
Built-in Prompt IDE for iterative prompt optimization
Self-hosted deployment on local, Docker, Kubernetes
Phoenix Cloud managed hosting option
Vendor-agnostic support for any model/framework
Native OpenTelemetry integration
OpenInference specification for LLM telemetry
Human annotation and automated labeling
Ghost trajectories to simulate alternative agent paths
Eval-as-you-test for early quality feedback
One-click integration with LlamaIndex
