Memory
Dynamo remembers things across sessions. The model saves important facts automatically, or you can ask it to remember something specific.
Memory is the model's own gitignored notebook — distinct from the committed, team-shared brief in `DYNAMO.md`. Memory is never auto-promoted into DYNAMO.md; see DYNAMO.md vs Memory for how the two systems divide up.
Memory Tools
| Tool | Description |
|---|---|
memory_save | Save a fact, preference, or project note |
memory_list | List all saved memories |
memory_read | Fetch the full content of one memory |
memory_delete | Remove a memory |
Scopes
Memories are stored in two locations:
| Scope | Location | Use for |
|---|---|---|
| Global | ~/.config/dynamo/memory/ | Your preferences, feedback, references |
| Project | .dynamo/memory/ | Project-specific context |
Browsing Memories
/memory # list all saved memories (global + project)
/memory consolidate # merge duplicates and clean up stale entriesHow It Works
The model proactively saves memories when it learns lasting facts — your role, preferences, project decisions, and important context. Each memory is a markdown file with frontmatter (name, description, type).
Memory types:
- user — your role, preferences, knowledge
- feedback — guidance on how to approach work
- project — ongoing work, goals, decisions
- reference — pointers to external resources
Consolidation
Over time, memories can accumulate duplicates or become stale. Run /memory consolidate to trigger AutoDream-style consolidation — the model merges duplicates, resolves contradictions, and prunes outdated entries.
Manual /memory consolidate works on any plan. Automatic consolidation on session start requires ai.auto_consolidate: true in dynamo.yaml (off by default) and a paid plan — the dream-on-startup loop is a paid feature.
To disable memory entirely: ai.memory: false in dynamo.yaml.