Note: For installation, use
deploy/install-claude-code.md— it handles everything including appending this template automatically.
This file contains the raw instruction block for manual reference. Copy the block below into ~/.claude/CLAUDE.md (global) or your project's CLAUDE.md (per-repo) if you prefer manual setup.
This is the Claude Code–specific block. It uses session-recall-cc, which reads Claude Code's own JSONL session transcripts (~/.claude/projects/) via a dedicated SQLite+FTS5 index — not the plain session-recall command, which reads Copilot CLI's session store instead. If your CLAUDE.md also contains the generic session-recall block (e.g. from copilot-instructions-template.md or deploy/install.md), make sure that block is gated to skip itself for Claude Code — see deploy/install-claude-code.md for details.
Requires auto-memory[claude] installed and SESSION_RECALL_ENABLE_CLAUDE_BACKEND=1 set.
For Claude Code only — if you are not Claude Code (e.g., Copilot CLI, Cursor, Aider), ignore this section.
When starting a new conversation in this repo, run:
session-recall-cc list --json --limit 5
Use the output to ground your understanding of recent work.
Searching past sessions:
session-recall-cc search "natural language phrase" # FTS5 full-text
session-recall-cc files # files touched recently
session-recall-cc show <session-id> # full transcript
Query tips: Multi-word natural-language phrases work well. Avoid hyphens / dots in single-token searches — FTS5 splits on them. Prefer
search "session recall"oversearch "session-recall". For exact filenames, usefiles | grep <name>instead ofsearch.
This requires auto-memory[claude] installed and SESSION_RECALL_ENABLE_CLAUDE_BACKEND=1 set.