Getting started
Three steps to start using memtomem.
Welcome to memtomem
Search your indexed memories by meaning — not just exact keywords.
Sources are the files you add. Chunks are the searchable pieces. Memories are everything together.
Select a result to view details
Pick a section in the sidebar to import what you already have, or create something new — then sync it out to Claude Code, Codex, and your other detected runtimes.
Context Gateway
iBrowse, switch, rename, and unregister the projects memtomem tracks. Health shows which roots are missing or not yet initialized.
Projects
Reusable workflows stored in .memtomem/skills/ and synced to each runtime's skills folder (.claude/skills/ etc.).
Skills
Project slash commands stored in .memtomem/commands/ and synced to each runtime's commands folder (.claude/commands/ etc.).
Commands
Specialized subagents stored in .memtomem/agents/ and synced to each runtime's agents folder (.claude/agents/ etc.).
Subagents
MCP server definitions stored in .memtomem/mcp-servers/ and synced to the project's .mcp.json. Project-scoped only — a single store, with no user-level copy.
MCP Servers i
Browse the global wiki (~/.memtomem-wiki) — a separate git library, not your ~/.memtomem user store or config — of canonical skills, subagents, and commands. Install snapshots an item into the current project's .memtomem/ store.
Wiki
i Global wikiLifecycle hooks run a command on events like a tool call or session end, stored in .memtomem/settings.json and applied to the active runtime only.
Hooks
One-shot scan. For automatic re-indexing, use Sources → Add path.
Advanced options
| Files scanned | |
| Total chunks | |
| Indexed | |
| Skipped | |
| Deleted | |
| Duration | |
| Namespace | |
| Blocked (secrets) | |
| Errors |
Drop files here or
.md .txt .json .yaml .yml .toml
Saved in
~/.memtomem/uploads/
·
Privacy guard scans for tokens, API keys, and private keys — not email or phone numbers.
Configuration
Server settings: embedding provider, storage backend, search and indexing parameters.
Configuration Guide
Select a config section to see its documentation.
Organize memories by namespace and review chunk counts per group.
Namespaces
What do these namespace labels mean?
Namespaces label where each memory came from so you can scope a search by origin. They are memtomem indexing labels — they do not replace each tool's own memory model.
default- Memories you add directly, with no namespace set.
claude:<project>- Claude Code per-project auto-memory, bucketed by the project folder above memory/.
codex:*- Codex memories, split into codex:rollout_summaries (session recaps), codex:extensions (ad-hoc notes), and codex:global (everything else).
gemini-memory:<slug>- Antigravity / Gemini global memory (~/.gemini/GEMINI.md), imported one-shot via mm ingest gemini-memory.
agent-runtime:<id>- A single agent's isolated scope during multi-agent runs.
shared- Knowledge shared across agents.
Mapping rules come from mm init or the Config tab. Learn more
Find near-duplicate chunks by similarity threshold and merge or skip them.
Scan previews candidates only. Merge buttons appear on reviewed pairs and delete only the duplicate chunk you choose.
Expire chunks older than a cutoff date, optionally filtered by source path.
Run Scan first. Expire stays disabled until the current filters have a reviewed candidate count.
| Total chunks | |
| Expired candidates | |
| Deleted chunks |
Permanently delete all memory data and reinitialize the database. Embedding config is kept.
This will permanently delete ALL data (chunks, sessions, history, relations, entities, etc.) and reinitialize the database. Embedding configuration is preserved.
Export a JSON bundle first if you may need to restore this database later.
Browse episodic session history and per-session event timelines.
Episodic Sessions
Session Events
Ephemeral key-value scratch pad with optional TTL, promotable to long-term memory.
Working Memory
Saved procedures (recipes) available for recall.
Procedures
Memory health report — coverage, dead chunks, session stats. Includes watchdog run.
Memory Health
Watchdog
Periodic background checks. Run manually to force an update.
Process-lifetime counts from the secret-redaction gate — how many writes were passed, blocked, or bypassed (Index without privacy gate), per surface. Resets on restart.
Redaction Stats
Export filtered chunks to a JSON bundle or import a bundle to re-index.
Export
Preview the filters before downloading so the bundle size is explicit.
Import
Upload a memtomem_export.json bundle to re-index its chunks.
| Total chunks | |
| Imported | |
| Skipped | |
| Failed |
Timeline
Review memory activity over time.