Multi-provider agent orchestration

Run coding agents from Claude, GPT, and Gemini side by side. Route tasks to the best model for each job with automatic provider selection.

New

You can now orchestrate coding agents across multiple providers simultaneously. Each task gets routed to the model best suited for the job, whether it is a complex refactor, a quick bug fix, or a code review.

Auto-routing analyzes your prompt and picks the optimal provider based on task complexity, context length, and model strengths. Override it anytime by pinning a specific model.

Improvements

Agent sessions now persist across app restarts with full context recovery
Token usage dashboard shows cost breakdown per provider and session
Streaming responses are 40% faster with chunked transfer encoding
Added keyboard shortcut (Cmd+K) to quick-switch between active agents

Fixes

Fixed token counting discrepancy with streaming responses from Anthropic models
Fixed dark mode rendering in the diff viewer
Resolved an issue where agent context could be lost during provider failover
Fixed clipboard paste not working in the agent input on Windows