CLI reference

The complete reference for the standalone remnic command-line interface. Every command here is part of the @remnic/cli package and runs against your local memory store — no account, no cloud. For the OpenClaw-hosted openclaw engram surface, see OpenClaw-hosted commands at the end.

Install

npm install -g @remnic/cli

This installs two binaries: remnic (canonical) and engram (a legacy forwarder kept during the rename window — every command below works under either name). You can pin the binary the auto-runner uses with REMNIC_CLI_BIN (or the legacy ENGRAM_CLI_BIN).

New to Remnic? Start with the quickstart, then come back here for the full command surface.

Conventions

• Most commands accept --json for machine-readable output.
• Configuration is resolved in this order:
  1. REMNIC_CONFIG_PATH environment variable (ENGRAM_CONFIG_PATH is still honored during the compatibility window)
  2. ./remnic.config.json in the current directory
  3. ~/.config/remnic/config.json (default)
• Placeholders use <angle-brackets>. Optional arguments use [square-brackets].
• Run any command with --help for its full option list.

Remnic ships 35 top-level command names (benchmark is a compatibility alias of bench), grouped below by what you reach for them.

Setup & health

Command	What it does
remnic init	Create remnic.config.json in the current directory
remnic migrate [--rollback] [--json]	Run — or with --rollback, undo — the first-run Engram-to-Remnic migration
remnic status [--json]	Show server/daemon status and health
remnic doctor	Diagnose the install (Node version, config, API key, memory dir, daemon)
remnic config	Print the resolved configuration
remnic daemon <start|stop|restart|install|uninstall|status>	Manage the background HTTP server. install/uninstall register or remove the OS service; start/stop/restart/status control a running one
remnic token <generate|list|revoke> [connector-id]	Manage bearer tokens for the access server
remnic onboard [dir] [--json]	Analyze a project directory (language detection, doc discovery, ingestion plan)

remnic init
remnic daemon start
remnic status
remnic doctor

Daily use

Command	What it does
remnic query <text> [--json] [--explain]	Query memories. --explain adds a per-tier latency and source breakdown
remnic xray <query> [--format text|markdown|json] [--budget <chars>] [--namespace <ns>] [--out <path>]	Run a recall with X-ray capture and print the unified tier + audit + MMR + filter snapshot
remnic briefing [--since <window>] [--focus <filter>] [--save] [--format markdown|json]	Generate a daily context briefing
remnic review <list|approve|dismiss|flag> [id]	Manage the review inbox
remnic curate <path> [--json]	Curate files into memory with duplicate and contradiction detection
remnic sync <run|watch> [--source <dir>]	Diff-aware filesystem sync into memory
remnic dedup [--json]	Find duplicate memories
remnic tree <generate|watch|validate> [--output <path>] [--categories <list>] [--max-per-category <n>] [--no-entities] [--no-questions] [--json]	Build and validate the workspace context tree

remnic query is the everyday entrypoint. Use --explain to see where a slow answer spends its time:

remnic query "what did we decide about the pricing model?" --explain

remnic briefing windows accept yesterday, today, NNh, NNd, or NNw; focus filters accept person:<name>, project:<name>, or topic:<name>:

remnic briefing --since 7d --focus project:acme-webshop --format markdown --save

Connectors & integrations

Command	What it does
remnic connectors <list|install|remove|doctor|marketplace|status|run> [id]	Manage host adapters and live connectors
remnic openclaw <install|upgrade|migrate-engram>	Wire Remnic into OpenClaw, upgrade the plugin, or migrate a legacy openclaw-engram install
remnic oauth <pending|approve|deny> [--format <fmt>] [--yes]	Manage pending OAuth authorizations (used by the ChatGPT MCP connector)
remnic wearables <status|check|sync|transcript|search|memories|speakers|corrections|fuse|fused>	Pull, clean, and store wearable transcripts (Limitless / Bee / Omi)
remnic location <status|check|sync|backfill|day>	Location day sync from registered providers (issue #2047; e.g. a self-hosted Reitti instance)

connectors install <id> writes the config a host tool needs to talk to Remnic. Built-in adapter ids include claude-code, codex-cli, cursor, cline, github-copilot, roo-code, windsurf, amp, pi, omp, replit, generic-mcp, weclone, and hermes. connectors status and connectors run operate on live connectors (background sync sources) rather than host adapters. connectors marketplace <generate|validate|install> manages Codex marketplace manifests.

remnic connectors list
remnic connectors install claude-code
remnic connectors status

remnic openclaw subcommands take the usual safety flags: --yes/--force to skip prompts, --dry-run to preview, plus --memory-dir, --config, --version, --plugin-dir, --legacy-plugin-dir, and --no-restart.

remnic wearables speakers has its own list|self|set|remove actions and corrections has list|add|remove; fuse/fused merge and read cross-source fused transcripts for a day. Each source ships as an à la carte connector package (for example npm install @remnic/connector-limitless).

remnic location runs the same shared runner as the MCP/HTTP location surfaces. Credentials for built-in providers come from the environment (REITTI_BASE_URL, REITTI_TOKEN, optional REITTI_AUTH_MODE=x-api-token|bearer); an absent provider package reports provider-not-registered instead of failing. sync covers location.syncDays ending yesterday (or --date/--days), backfill --from --to is capped at 90 days, and coordinates are never stored or printed unless location.retainCoordinates is enabled.

Memory management

Command	What it does
remnic space <list|switch|create|delete|push|pull|share|promote|audit>	Manage personal, project, and team memory spaces. create accepts --parent <id>
remnic versions <list|show|diff|revert> <page-path> [id] [--json]	Page-level version history: list, show, diff, or revert snapshots
remnic taxonomy <show|resolver|add|remove|resolve>	Manage the MECE knowledge-directory taxonomy. add <id> <name>, remove <id>, resolve <text> [--category <cat>]
remnic enrich <entity-name|--all|--dry-run|audit|providers>	Run the external entity-enrichment pipeline
remnic procedural stats [--format json|text] [--memory-dir <path>]	Print procedural-memory stats (counts, recency, config)
remnic capsule <fork|lineage>	Fork a portable capsule archive into a memory root, or print a fork's lineage breadcrumb
remnic offline <prepare|sync|status|watch>	Remote/offline memory sync
remnic action-confidence [--action <text>] [--confidence <0-1>] [--risk <level>] [--context <readiness>]	Evaluate the read-only ask/draft/act/refuse/escalate advisory policy

remnic capsule is the portability entrypoint:

remnic capsule fork <archive.capsule.json.gz> --target <memory-dir> --fork-id <id>
remnic capsule lineage --fork-id <id> --root <memory-dir>

Import & export

Command	What it does
remnic import --adapter <name> --file <path> [--dry-run] [--batch-size <n>]	Import from a supported export file
remnic import-lossless-claw --src <path> [--dry-run] [--session-filter <id>]	Migrate a lossless-claw LCM database into Remnic's LCM mode
remnic training:export --format <name> --output <path>	Export memories as a fine-tuning dataset

remnic import --adapter supports exactly these adapters: chatgpt, claude, gemini, mem0, and supermemory. WeClone is not an import adapter — it is hosted-only via openclaw engram bulk-import --source weclone. Lossless-claw has its own dedicated remnic import-lossless-claw command above.

remnic import --adapter chatgpt --file ./conversations.json --dry-run

Advanced & research

Command	What it does
remnic bench <action> [benchmark...]	Run and manage benchmarks (alias: remnic benchmark)
remnic binary <scan|status|run|clean>	Binary-file lifecycle: scan, status, run [--dry-run], clean --force
remnic extensions <list|show|validate|reload>	Manage memory extensions

bench actions: list, run, datasets (download/status), runs (list/show/delete), compare, results, baseline (save/list), export, publish, published, judge-calibrate, ui, providers (discover). The check and report actions run under the benchmark alias. published and judge-calibrate are maintainer tools for the published-benchmark harness.

remnic bench list
remnic bench run --quick

remnic extensions reload is a reserved no-op (extension caching is not yet implemented); the other actions are live. See the benchmarking guide for the full bench workflow.

OpenClaw-hosted commands

When Remnic runs inside OpenClaw, the gateway registers a single top-level engram command whose children are the deeper memory-operations surface — roughly 100 subcommands invoked as openclaw engram <command>. This surface is intentionally separate from the standalone remnic binary: it is the operator and maintenance toolkit that runs against the gateway's live orchestrator. There is no openclaw remnic namespace; inside a chat session the plugin exposes the /remnic <on|off|status|clear|stats|flush> session command instead.

Most-used openclaw engram commands:

openclaw engram stats                 # memory counts, buffer state, QMD status
openclaw engram search "<query>"      # search memories from the terminal
openclaw engram recall "<query>"      # run a full recall
openclaw engram doctor                # diagnose the hosted install
openclaw engram export                # export memories to a portable file
openclaw engram import                # import from a portable file
openclaw engram backup                # snapshot the memory directory
openclaw engram purge                 # delete memories (guarded)
openclaw engram access http-serve     # start the HTTP access surface
openclaw engram access mcp-serve      # start the MCP access surface
openclaw engram access-stats          # most-accessed memories report
openclaw engram bulk-import --source weclone

Note the distinction: access is the group that manages the HTTP and MCP access surfaces (http-serve, http-stop, http-status, mcp-serve), while access-stats is the separate most-accessed-memories report.

Beyond these, the hosted surface covers namespaces, capsules, trust zones, dreams, peers, secure-store, governance, review, continuity, the work layer (task/project), and the research and creation-ledger command families. The operations guide documents the operator workflows, and API reference — CLI commands is the full openclaw engram command table.