Retrace a failure
Move from a short summary back to the prompt, output, and run context behind it.
Run history for AI-built software
BitterLog keeps the original history behind an agent run, so you can retrace a failure, review a decision, and continue the work without relying on a summary alone.
Prompts, output, timestamps, and human notes stay connected to the run that produced them.
Already use Bitter? Sign in to continue.
Move from a short summary back to the prompt, output, and run context behind it.
Keep the original record and the human note explaining what changed side by side.
Give the next review a reliable starting point instead of asking memory to fill the gaps.
Why BitterLog
Agent summaries are designed to compress. That is useful until a failure or handoff depends on a detail the summary left out. BitterLog keeps a path back to the source without asking you to reconstruct the run from memory.
How it works
Send events or raw session files through the CLI or API. BitterLog preserves the source before anything is summarized.
Use the run identity to retrieve the original object, including its timestamp and stable content fingerprint.
Place annotations beside the retained record so later reviews include both what happened and why the work changed.
Developer surface
Write and retrieve run history through the Bitter CLI or API. The public index and status page show which read paths are available today.
bitter log put --bucket agent_runs --run-id 5468 run.jsonl
bitter log list --bucket agent_runs --run-id 5468
bitter log get --bucket agent_runs --id 71 --body-onlyAccess
BitterLog access starts through Bitter. During launch, evidence onboarding remains request-reviewed.
Prefer email? founders@bitterlog.com