ROADMAP / EVIDENCE BEFORE LABELS
The roadmap advances through verifiable capability.
A live page is not the same as a finished product. Each stage requires contract, end-to-end, or production evidence, with external gaps stated explicitly.
Real Agent user loop
Email identity, workspaces, encrypted Bailian credentials, qwen3.8-max Agent Runs, SSE, cancel and retry, Skills, artifacts, usage, audit, and public MCP are in production.
Live sandbox providers
ACS and E2B adapters pass offline contracts, but billable live-environment smoke tests remain. The next stage verifies create, command, file, cleanup, recovery, and cost boundaries.
Independent failure domains
Complete independent-host Redis HA, multi-hour soak, external on-call delivery, and signed immutable images. These gates remain open until real evidence exists.
Benchmarks, more providers, and self-service
Sandbox benchmarks, additional model and execution providers, enterprise identity, and self-service plans follow stable live-provider data.Service statusChangelog