🔍 Sherlock Live Inspector

Connect any LLM — cloud or a local OpenAI-compatible server — and watch Sherlock curate its context in real time.

Providers — connect at least one; mix providers per role

all roles

Main · LLM-1

Summarizer · LLM-2

Inferencer · LLM-3

Visualizer · LLM-4

System prompt (LLM-1 persona)

🔬 Companions Background Redact secrets 🔎 Verify

turbo runs LLM-2 + LLM-3 every turn, cold_start wakes them only when a real signal calls for it, and off stays single-model.

🧠 Long-term memory profile incognito — read, don't save

Durable facts (name, allergies, “remember this”) are saved to the chosen profile and survive restarts; incognito keeps recalling them but pauses new writes.

📊 Visualizations

Where a chart genuinely helps, LLM-1 proposes one and LLM-4 renders it inline as a sandboxed artifact.

🌐 Web search

LLM-1 search and LLM-3 freshness checks share this engine; with a key, Brave/Tavily/Valyu beat the free DuckDuckGo by a wide margin.