buildingopen

Session Recall — Development skill for Claude Code

Development community

Search and analyze Claude Code session transcripts to recover context after compaction.

How to install Session Recall

This entry records only its repository, not the path inside it, so there is no exact command to give. Open buildingopen/session-recall and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Session Recall does

Search and analyze Claude Code session transcripts to recover context after compaction.

Alternatives in Development

Ccr — CCR (Compressed Context with Retrieval) — recall the full original of context that great_cto compressed/filter 83 ★
Fire Cost — Track context usage, estimate session costs, and trigger strategic compaction 77 ★
Context Pressure — Diagnose long-search, large-file, and context-window pressure in a live session 58 ★

README

session-recall

[![npm version](https://img.shields.io/npm/v/session-recall)](https://www.npmjs.com/package/session-recall) [![license](https://img.shields.io/npm/l/session-recall)](https://github.com/buildingopen/session-recall/blob/main/LICENSE) [![npm downloads](https://img.shields.io/npm/dm/session-recall)](https://www.npmjs.com/package/session-recall)

Total recall for Claude Code. Get back everything compaction took.

npx session-recall "the thing you lost"

The problem

When Claude Code compacts context, you get a vague summary. Details vanish: decisions, error solutions, specific commands that worked, corrections you gave.

But the raw transcripts are never deleted. Every session is saved as JSONL in `~/.claude/projects/`, permanently. Compaction only affects the live context window. The data is always there, just buried in JSON soup.

session-recall parses those JSONL transcripts properly. It extracts human-readable messages, tool calls, errors, and patterns, so you find what you need in seconds.

Search any past session

Find exactly what was lost after compaction:

Analyze your sessions

`--report` finds retry loops, errors, user corrections, inflated self-scores, and generates CLAUDE.md rules to prevent the same mistakes:

session-recall report demo

Cross-session patterns

`--all` analyzes your last N sessions to find recurring problems:

$ session-recall --all 10

CROSS-SESSION SUMMARY (10 sessions)
  Total tool calls: 4351 | Total errors: 165

RECURRING RETRY PATTERNS
  Bash: retried in 7/10 sessions (avg 4.2x when it happens)

SELF-SCORING ACCURACY
  185 scores across 10 sessions (avg 7.7/10)
  20/185 (11%) had user issues after

RECURRING ERROR TYPES
  COMMAND_FAILED: in 8/10 sessions
  FILE_NOT_FOUND: in 6/10 sessions