REDL
REDL.IO
We've fully migrated REDL's AI engineer from Google Gemini to Claude by Anthropic — faster, smarter, and cheaper for you.
Per-server model selection
A version switcher now appears below the input field in the AI chat:
Next to it — an effort toggle (Low / Medium / High / Max): the higher it is, the more carefully the agent thinks (for Sonnet/Opus).
Memory
The agent remembers your server's context between sessions (Claude's own session + the
Smart context compression
Previously, context would compress too early and the chat would "freeze up." Now Claude compresses cleanly only at the limit — no more lockups. We've also added a "Compress" button at the top of the chat for manual compression at any time (with memory preserved).
Clean responses + "Work Log"
Intermediate agent steps ("downloading…", "checking logs…") no longer clutter the final response — they're collapsed into an expandable "Work Log" block beneath the answer.
Precise cost tracking
AI budget usage is calculated directly from the actual cost of Claude requests (with caching factored in — which saves your budget even further).
Post questions and feedback in this thread
What's New
Per-server model selectionA version switcher now appears below the input field in the AI chat:
- Claude Haiku 4.5 — fast and cheap (default)
- Claude Sonnet 4.6 — balanced speed and intelligence
- Claude Opus 4.7 — maximum power for complex tasks
Next to it — an effort toggle (Low / Medium / High / Max): the higher it is, the more carefully the agent thinks (for Sonnet/Opus).
MemoryThe agent remembers your server's context between sessions (Claude's own session + the
CLAUDE.md project file). No more losing track of the conversation. Smart context compressionPreviously, context would compress too early and the chat would "freeze up." Now Claude compresses cleanly only at the limit — no more lockups. We've also added a "Compress" button at the top of the chat for manual compression at any time (with memory preserved).
Clean responses + "Work Log"Intermediate agent steps ("downloading…", "checking logs…") no longer clutter the final response — they're collapsed into an expandable "Work Log" block beneath the answer.
Precise cost trackingAI budget usage is calculated directly from the actual cost of Claude requests (with caching factored in — which saves your budget even further).
How to use it
Open your server → the AI tab → describe your task in plain language (e.g., "install EssentialsX and restart"). Need a smarter result — switch to Sonnet or Opus and raise the effort. Everything is already live on your servers.Post questions and feedback in this thread
