Your AI agents have hundreds of conversations. Decisions get made, patterns get discovered, gotchas get found — and then they’re lost in chat history. hub consolidate fixes that.

What’s new

hub consolidate

Reads chat sessions from all your editors, compacts them into a batch, and spawns your editor’s CLI to extract useful knowledge into ./memories/.

hub consolidate                    # Process last 20 sessions
hub consolidate --last 50          # Last 50 sessions
hub consolidate --since 2026-03-01 # Since a specific date
hub consolidate --editor kiro      # Only from Kiro
hub consolidate --cli claude       # Use Claude CLI for extraction
hub consolidate --dry-run          # Preview without extracting
hub consolidate --reset            # Reprocess all sessions

How it works

1. Reads chat history from Kiro, Claude Code, and OpenCode (auto-detected)
2. Normalizes messages across editors into a unified format
3. Sorts by most recent first, skips already-indexed sessions
4. Writes a compacted batch to .hub/consolidation/
5. Spawns the editor CLI (kiro-cli, claude, or opencode) with a knowledge extraction prompt
6. The agent reads the batch, checks existing memories for duplicates, and writes new .md files to ./memories/{category}/
7. Marks processed sessions in .hub/consolidation-state.json

Supported editors

Zero extra dependencies

No API keys, no SDK, no external LLM service. It uses the CLI of whatever editor you already have installed — same model, same subscription, same config. The pattern is identical to how agent-teams-lead spawns teammates.

Scheduling

Run it automatically with cron, launchd, or systemd. See the Memory docs for setup instructions per OS.

Bug fixes

• Fixed session ordering: collectors now sort globally by date before applying the limit, ensuring the most recent sessions are always processed first regardless of which workspace directory they’re in.

Upgrade

npx @arvoretech/hub@0.16.0 generate