Claude Code subscribers get a usage bar, but little insight into what’s driving it. Paweł Huryn built a lightweight local dashboard that reads Claude Code’s JSONL logs, stores them in SQLite, and charts tokens, models, time ranges, and estimated costs.
Anthropic has rolled out an advisor tool for Claude that lets Sonnet or Haiku run tasks and consult Opus only at key decision points. The beta promises better agent performance while keeping costs down, all within a single API request.
Codex has just rolled out a $100 Pro plan, positioned as a new middle tier for heavier users. Tibo Sottiaux also says usage limits are being reset again, and the 2X limits promotion on $100 and $200 Pro is being extended.
Anthropic has just rolled out Claude Managed Agents, bringing a hosted runtime and tools to run AI agents in production. The public beta adds sandboxed execution, tracing, long-running sessions, and multi-agent coordination. Pricing starts at $0.08 per session-hour plus tokens.
GoogleCloudPlatform has published Scion, an experimental orchestration testbed for running multiple AI coding agents in parallel. Each agent runs in an isolated container with its own git worktree, credentials, tmux control, and OTEL telemetry.