tuneloop./docs
GitHub

Session Transcripts

Tuneloop captures every AI coding session and enriches it with structured metadata. Search and filter sessions across all harnesses in one place.

Tuneloop sessions list — searchable table with filters for work type, repo, outcomes, date range, and tags, showing session title, cost, outcomes, complexity, and work type

Session Enrichment

Each session is automatically enriched with:

  • Outcome links — merged PRs, features shipped, files changed
  • Cost attribution — granular cost attributed to specific outcomes
  • Complexity — task complexity classification (trivial, routine, substantial, open-ended)
  • Autonomy — how much the agent did independently vs. with user guidance
  • Work type — implement, debug, refactor, review, docs
  • Key decisions — significant choices the agent made during the session
  • Tool error categories — classified errors for every tool call

Transcript View

Each session is rendered as a turn-by-turn conversation between the user and the agent, with:

  • Tool calls shown inline with collapsible diffs, commands, and outputs
  • Error panels highlighting tool failures with error navigation
  • Subagent transcripts as separate scoped views
Tuneloop session transcript — turn-by-turn view with inline tool calls, error highlighting, error navigation, subagent threads, and work type block filter

Block Filter

Use the "View by" filter to focus the transcript on a single PR, feature, or work type. Other content is collapsed behind named gap markers, letting you trace exactly what work went into a specific outcome.

Files Tab

See every file the agent touched, with:

  • Per-file diff cards grouped by which user message drove the edits
  • Running-state diffs showing the file's evolution through the session
  • Narration captions explaining each change

Summary Tab

A structured overview of the session:

  • All facet dimensions (work type, complexity, autonomy)
  • Intent summary and key decisions
  • Features, PRs, and tickets touched
  • Files modified
  • Outcomes produced

Share Sessions

By default, sessions are visible only to the creator. Tuneloop allows users to share sessions with others for debugging — share links can be posted to PR reviews, for example.

Skills Health