Persistent cross-session memory for Claude Code, powered by ZeroDB.
Claude Code loses all context at the end of every session. This plugin fixes that. It automatically stores what Claude learns during a session and recalls it the next time you start working — architecture decisions, bug discoveries, conventions, code ownership, and in-progress work.
# From the official Claude Code marketplace
/plugin install zerodb-memory@claude-plugins-official
# Direct from GitHub (no marketplace needed)
/plugin install github:AINative-Studio/zerodb-claude-pluginNothing to do. On first use the plugin auto-provisions a free trial ZeroDB project for you and stores the credentials at ~/.zerodb/credentials.env — no signup, no exported env vars, no copy-pasted keys. The trial is unclaimed by default; the plugin surfaces a claim link the first time this happens so you can attach it to a real account and keep it permanently.
Already have a ZeroDB account? Set these before your first session and the plugin will use them instead of provisioning a trial:
export ZERODB_API_KEY=your_api_key_here
export ZERODB_PROJECT_ID=your_project_id # optional, auto-detected if omittedTo disable auto-provisioning entirely (e.g. in CI or an offline environment), set ZERODB_NO_AUTO_PROVISION=1.
At session start: Claude automatically loads memories for the current project and announces what was recalled.
At session end: Claude extracts key facts from the conversation — decisions made, bugs found, conventions established — and stores them in ZeroDB.
Anytime:
/remember Sarah owns the payments module — get her sign-off before refactoring
/recall how do we connect to the database
/forget old staging URL
/memory stats| Command | Description |
|---|---|
/remember <text> |
Manually store a memory |
/recall <query> |
Search memories semantically |
/forget <query> |
Remove matching memories (with confirmation) |
/memory list |
Browse all memories for this project |
/memory stats |
Memory count, types, last sync |
/memory autopersist [on|off|review] |
Toggle auto-persist |
/memory autorecall [on|off] |
Toggle auto-recall |
/memory export |
Export memories as Markdown or JSON |
/memory clear |
Wipe all memories for this project |
| Env var | Default | Description |
|---|---|---|
ZERODB_API_KEY |
auto-provisioned | Your ZeroDB API key — set it to use an existing account instead of a trial |
ZERODB_PROJECT_ID |
auto | ZeroDB project ID (auto-detected from git remote) |
ZERODB_NO_AUTO_PROVISION |
unset | Set to 1 to disable trial auto-provisioning and fail instead |
ZERODB_AUTOPERSIST |
on |
Auto-persist mode: on, off, or review |
ZERODB_AUTORECALL |
on |
Auto-recall on session start: on or off |
- Memories are facts and observations — never raw file contents, secrets, or credentials
- Stored in your ZeroDB account, not shared with anyone
- Use
/memory clearto wipe all memories for a project at any time
Memory operations bill to your ZeroDB account:
| Operation | Credits |
|---|---|
| Store memory | 1 credit |
| Semantic search | 2 credits |
| Get context bundle | 3 credits |
Average session: 25 credits ($0.25). See pricing.
Full docs: docs.ainative.studio/claude-plugin
MIT — AINative Studio