LoopPrism

See what your coding agents actually do — every session, step by step. Local-first, zero setup.

Claude Code Codex Gemini CLI OpenCode DeepSeek Harness (dsh) WorkBuddy

Trajectory & Failure Flags

Inspect prompts, injected context, thinking, tool calls, durations, repeated loops, and failure streaks on one timeline.

Context & Timing Breakdown

See what filled the model input (system prompt, tool schemas, skills, messages, tool outputs), cache hit rate, and thinking vs. tool time.

Subagents & Token Spend

Trace main-agent to subagent branching flows and aggregate token cost priced at model list prices across apps.

LoopPrism preview
Full LoopPrism Tour (Chat, Trajectory, Context, Agents, and Cost — generated demo sessions)