Memory & Context

Hand the current conversation to a fresh background run that continues the work

Compact the current conversation into a handoff note for another session

Save and resume per-project context across Claude Code sessions with simple commands

Check how much context budget is left and how much installed skills consume

Legacy v1 stop hook that extracts session lessons, superseded by continuous learning v2

Learn project-scoped instincts from sessions and evolve them into reusable skills

Report Claude Code token use, spending and budgets by model, session or date

Write growth log entries that capture root causes and reusable rules from finished work

Refine context retrieval over several passes when a subagent lacks what it needs

Compact context at task boundaries instead of relying on automatic compaction

Offer 25, 50, 75 or 100 percent answer depth with token estimates before answering

Hand off work state and context between coding agents through a shared local vault

Configure and troubleshoot Honcho long-term memory for agents

Build and query an interlinked Markdown knowledge base with the LLM Wiki pattern

Give an agent long-term memory and retrieval with Pinecone

Pull standup summaries and usage tips from your Copilot session history

Restore working context saved earlier with the context-save command

Save the current working context so a later session can pick it up

Record, review and manage lessons learned from past coding work

Install gbrain and set up a local PGLite or Supabase memory store for coding

Re-sync gbrain with the current code and refresh search notes in CLAUDE.md

Answer codebase questions from a persistent knowledge graph of code and docs

Compress memory and todo files into caveman format to save input tokens with a backup

Apply cost-lowering fixes from a learn report with consent for each edit

Show output and cache token usage for the current session and its mode

Rebuild an LLM wiki index from the actual page tree and reconcile missing entries

Maintain a cited local company wiki with the LLM Wiki plugin tools

Turn project activity bundles into wiki standups, project pages and decision records

Store and organize durable knowledge across sessions with a file-based PARA memory

Ingest raw sources into an LLM wiki as source, entity, concept and synthesis pages

Audit an LLM wiki for contradictions, orphans, broken links and index drift

Answer questions from wiki pages and raw sources with citations and named gaps

Run an hourly claude-mem cycle that checks worker health and caches observations

Install claude-mem for Cursor with a local or remote worker and observer

Set up or check cloud sync and backup of your memories with cmem.ai Pro

Write a HANDOFF.md so a fresh session can pick up exactly where you stopped

Reply to claude-mem observer jobs with a skip summary or one observation XML block

Explain how claude-mem captures observations and where its memory data lives

Build and query knowledge bases from your claude-mem session observations

Search past sessions and memories for earlier decisions, files and tasks

Pair the claude-mem Cowork plugin with your cmem.ai account and sync token

Create and verify custom claude-mem modes that control what gets remembered

Search your Screenpipe screen and audio history and save key context

Search code structure with tree-sitter to save tokens over reading full files

Configure and test prompt compression rules to cut token use before requests

Manage context configs, RTK filter sets and sessions for OmniRoute from the CLI

Manage the LLM response cache, TTLs and semantic similarity thresholds

Configure RTK and Caveman prompt compression to cut token use

Set RTK filters and context rules and test them on real prompt samples