CLI guide

The CLI is the default way to use Memory. The day-to-day surface is small:

memory query "<question>"   # pull product context mid-task
memory save --stdin         # save product-meaningful changes
memory sync                 # reconcile at session end
memory status               # where things stand

Everything else is setup, inspection, or maintenance. All commands accept a global --json flag for structured envelopes.

The loop

`memory init`

Initialize Memory storage in this project and print the indexing brief.

memory init
memory init --dry-run
memory init --no-view
memory init --no-agent-guidance
memory init --force
memory init --brief

Creates .memory/ storage with a starter project node.
Installs the guidance block and generated product map sections into AGENTS.md and CLAUDE.md (skipped with --no-agent-guidance).
Starts the local viewer (skipped with --no-view).
Prints the indexing brief the coding agent follows to build the initial graph. --brief prints only the brief and touches nothing.
--dry-run previews what init would create or change without writing.
--force discards existing Memory storage and initializes from scratch.

`memory query <question>`

Query local Memory and print a token-budgeted Markdown subgraph of matching memory.

memory query "why do webhook retries run in the worker?"
memory query "state of batch exports" --budget 1200

Seeds on full-text matches, expands one hop along active relations, and attaches connected open questions.
--budget <number> overrides the token budget (default comes from .memory/config.json, initially 2000).

`memory save --stdin`

Save product memory from intent-first input: create or update feature/decision/gotcha/question nodes, mark stale, supersede, or delete.

memory save --stdin
memory save --stdin --dry-run

--stdin is required; the input is JSON with the shape {task, nodes, stale, supersede, delete}. See the Reference for the exact node fields.
--dry-run validates and plans the write without changing anything.
Every successful save refreshes the product map in AGENTS.md/CLAUDE.md.

`memory sync`

Run the diff-driven staleness pass: report nodes whose anchors changed or died since the last sync, list coverage gaps, refresh the product map, and advance the sync marker.

memory sync
memory sync --dry-run

Diffs the tree since the commit recorded in .memory/sync-state.json.
Reports changed, orphaned, and unanchored nodes plus directories with code but no feature coverage, and prints an agent prompt with a pre-filled save skeleton when reconciliation is needed.
Never writes graph nodes itself. --dry-run reports without advancing the marker or refreshing the map.

`memory status`

Summarize the product graph: features by stage, open questions, stale anchors, last activity, and last sync.

memory status
memory status --all

--all prints one row per registered project from the user-level registry (~/.memory/projects.json) — the cross-project dashboard.

Inspection

`memory check`

Validate canonical storage and generated index health. Also warns when anchors match no files or the product map sections in AGENTS.md/CLAUDE.md are missing or out of date.

memory check
memory check --json

`memory diff`

Show Memory changes, including untracked memory files in Git projects. Plain git diff -- .memory/ can miss untracked files before staging.

memory diff

`memory inspect <id>`

Show one Memory object and its direct relations.

memory inspect feature.webhook-retry-queue

`memory view`

Start the local viewer: projects dashboard, memory list, node detail, and the relation graph. Binds to 127.0.0.1 only.

memory view
memory view --open
memory view --port 4888
memory view --detach

--open opens the URL in the default browser.
--port <number> picks the port (random available port by default).
--detach starts the viewer in a background process and prints its URL.

Maintenance

`memory rebuild`

Rebuild generated indexes from canonical storage. Does not change canonical memory.

memory rebuild

`memory projects`

Manage the user-level project registry behind memory status --all and the viewer’s projects dashboard.

memory projects list
memory projects add            # registers the current directory
memory projects add /path/to/project
memory projects remove <registry-id|project-id|path>
memory projects prune          # drop entries whose storage is gone

`memory reset`

Back up and clear local Memory storage. The backup archive lands under .memory/.backup/.

memory reset
memory reset --all       # every project in the registry
memory reset --destroy   # delete .memory/ entirely, no backup

`memory upgrade`

Removed as a migration path. Storage created by older schema versions is not migrated; run memory reset then memory init and rebuild the graph from the indexing brief.

`memory docs`

Read bundled public docs or open the hosted docs site.

memory docs
memory docs getting-started
memory docs cli --open
memory docs --json

MCP

MCP covers query, save, status, and inspect when the client has launched memory-mcp. Everything else — init, sync, viewer, registry, maintenance — stays in the CLI. See the MCP guide.