Memory & Context

Restore saved project context, decisions and work state when resuming interrupted work

Restore saved decisions and work state to resume interrupted sessions

Save project state and decisions to searchable notes for handoffs and later sessions

Manage context and memory across long-running multi-agent workflows with vector stores

Save Lean 4 proof patterns and project conventions to persistent memory

Understand how context works in agent systems, and what limits it

Curate lessons from reflections and critiques into a project instruction file

Draft team decisions and convention notes for review in a repo knowledge wiki

Prepare durable git-based handoffs so the next engineer has what they need

Restore the latest saved diagnosis snapshot to continue where you left off

Save the current diagnosis state to disk for pickup in a later session

Track long-running goals durably with progress notes an agent can resume

Review agent memory for contradictions, gaps and stale facts, then propose fixes

Save a structured state snapshot with plan, decisions and next steps before stopping

Save a structured session summary before auto-compaction removes context

Check how much context is used and how close auto-compact is

Review saved memory notes with a critic council and promote the strongest entries

Compress the current chat into a handoff document for the next agent to continue

Look up team and org knowledge stored in Hivemind alongside built-in memory

Search shared team and org memory in Hivemind when recalling past information

Resolve duplicate and conflicting definitions found while ingesting context bundles

Capture durable business definitions and conventions into a cross-linked wiki

Cut agent output tokens with a compressed reply style that keeps accuracy

Search and evolve a SQLite knowledge graph of facts, concepts and links

Find directories holding knowledge files through a fast SQLite index

Evolve a population of knowledge graphs for cross-domain speculative synthesis

Recall and manage directory-level memories and links stored in YAML files

Move agent memory notes into tool-neutral docs that other coding CLIs can read

Pick up interrupted coding agent sessions by reconstructing the unfinished next steps

Resume interrupted Codex runs by verifying history and rebuilding the next step

Search local conversation history across coding agent tools and route to the reader

Search and export local coding agent session history and verbatim user inputs

Read, search and export local Codex history without resuming any work

Organize project files and operating notes into facts for content generation

Import memory, prompts and chats from other AI assistants by mapping their exports

Import your ChatGPT conversation history so past chats carry over

Load a large dataset into memory as a skimmed map that points to raw files

Migrate stored assistant memory from version 1 to version 2 in a one-time run

Migrate memory v2 concepts into a v3 wiki, with review and retrieval evals

Audit and restructure a long-term memory tree for beliefs and retrieval paths

Resolve contradictions between memory nodes that cannot both stay true

Purge memory entries that never change behavior and have no real-world anchor

Fix memory disclosure text, parent placement and aliases so notes stay findable

Split oversized memory nodes that mix unrelated concepts into focused ones

Extract shared patterns from repeated lessons and recurring mistakes in memory

Build a shared project glossary and resolve renamed or deprecated domain terms

Save facts, caveats and small rationales to the memory store of a project

Keep multi-step tasks on track across context resets and long sessions

Turn a finished or failed work cycle into reusable lessons and quality gates