Signet CLI Reference

Complete reference for all Signet CLI commands. For the Daemon HTTP API, see Api. For initial setup walkthrough, see Quickstart.

Path note: $SIGNET_WORKSPACE means your active Signet workspace path. Default is ~/.agents, configurable via signet workspace set <path>.

Installation

# bun (recommended)
bun add -g signetai

# npm
npm install -g signetai

# or installer script
curl -sL https://signetai.sh/install | bash

Runtime operations that need package execution (skills, updates) follow the primary package manager captured during setup, with deterministic fallback when unavailable.

Commands Overview

signet (No Arguments)

Shows the top-level help output with examples. This keeps the CLI safe to call from scripts and agents without dropping into an interactive menu.

  ◈ signet v0.1.0
  own your agent. bring it anywhere.

  Usage: signet [options] [command]

  Examples:
    signet setup
    signet status
    signet doctor
    signet doctor hermes
    signet daemon start
    signet remember "Nicholai prefers command-first CLIs"

Use explicit commands for interactive flows:

• signet setup — initialize or migrate a workspace
• signet configure — edit agent settings interactively
• signet doctor — troubleshoot local issues
• signet doctor hermes — troubleshoot Hermes Agent integration issues

signet desktop

Builds the official Electron desktop app from an existing Signet source checkout. Without --repo or SIGNET_SOURCE_DIR, the command first syncs the managed workspace checkout at <workspace>/signetai, ignoring generated desktop build artifacts such as surfaces/desktop/release/ so stale local artifacts do not block fast-forward pulls. Explicit source paths are left under user control; set SIGNET_SOURCE_DIR or pass --repo <path>. To build the checkout you are currently in, run signet desktop install --repo ..

signet desktop build
signet desktop install
signet desktop install --repo .
signet desktop install --repo ~/signet/signetai
signet desktop install --skip-build

signet desktop install runs bun install, then bun run build:desktop, then installs the newest built artifact. Linux/Arch currently installs a user-level AppImage launcher at ~/.local/bin/signet-desktop and a desktop entry under ~/.local/share/applications/signet.desktop. macOS and Windows builds are produced by the desktop package, with native installer automation still guarded until those platform installers are wired.

signet setup

Interactive first-time setup wizard (with optional non-interactive mode). Creates the $SIGNET_WORKSPACE/ directory and all necessary files.

signet setup
signet setup --path /custom/path
signet setup --non-interactive \
  --name "My Agent" \
  --harness claude-code \
  --identity-preset minimal \
  --deployment-type vps \
  --embedding-provider native

Options:

Non-interactive behavior:

• setup method: create new identity (no GitHub import)
• provider flags are optional; setup infers defaults from --deployment-type
• --identity-preset defaults to minimal for new workspaces
• the Minimal preset creates AGENTS.md for startup context and DREAMING.md for dreaming sessions; DREAMING.md is not loaded into normal startup context
• hooks/connectors are configured for requested harnesses
• with --deployment-type vps, setup prefers non-local extraction defaults from selected harnesses when those tools are available locally, then other detected tooling (claude-code, codex, opencode), and falls back to none when needed
• for existing-identity migration, previously configured extraction providers are preserved unless --extraction-provider is explicitly passed
• the bundled Signet Secrets core plugin is enabled by default; pass --disable-signet-secrets to opt out while leaving it installed
• GraphIQ is optional and disabled by default; pass --with-graphiq to install it via the bundled install script (downloads from GitHub releases)
• explicit provider flags override inferred defaults
• git: enabled unless --skip-git is passed
• when OpenClaw points at this workspace and no origin remote exists, setup requires either backup creation (--create-local-backup) or explicit bypass (--allow-unprotected-workspace)
• snapshot-backed protection is treated as “fresh” for 7 days; after that, status/doctor warn again unless a remote origin exists or a new snapshot is made

Extraction safety note:

• intended usage is claude-code on Haiku, codex on gpt-5.4-mini with a Pro/Max subscription, or local ollama with at least qwen3:4b
• with --deployment-type vps, setup avoids defaulting to local ollama extraction and prefers non-local providers
• set --extraction-provider none on a VPS if you do not want background extraction
• remote API extraction can create extreme usage fees fast

Wizard steps:

1. Agent Name - What to call your agent
2. Harnesses - Which AI platforms you use:
   • Claude Code (Anthropic CLI)
   • Codex
   • OpenCode
   • OpenClaw
   • Oh My Pi
   • Pi
   • Hermes Agent
   • Forge
3. OpenClaw Workspace - Appears only when an existing OpenClaw config is detected; workspace is patched only if you opt in, and setup warns that uninstalling OpenClaw can delete this workspace unless backups exist
4. Description - Short agent description
5. Core Plugins - Signet Secrets explains encrypted local storage, value-safe CLI/MCP/SDK access, command injection with output redaction, and connections to Signet’s local encrypted store and compatible 1Password references, then asks whether to enable the bundled signet.secrets plugin
6. Optional Code Retrieval - GraphIQ explains fast local codebase indexing, structural context, constants, and blast-radius tools, then asks whether to install the verified managed signet.graphiq plugin
7. Deployment Context - Where Signet is running (local, vps, server) to show environment-aware guidance before extraction provider selection
8. Embedding Provider:
   • Built-in (recommended, no setup required)
   • Ollama (local)
   • OpenAI API
   • Skip embeddings
9. Embedding Model - Based on provider:
   • Built-in: nomic-embed-text-v1.5
   • Ollama: nomic-embed-text, all-minilm, mxbai-embed-large
   • OpenAI: text-embedding-3-small, text-embedding-3-large
   • Ollama selections run preflight checks for binary availability, service health, and model presence; if checks fail, setup offers retry, switch to built-in embeddings, switch to OpenAI, or continue without embeddings
10. Search Balance - Semantic vs keyword weighting
11. Advanced Settings (optional):

• top_k - Search candidates per source
• min_score - Minimum search score threshold
• session_budget - Context character limit
• decay_rate - Memory importance decay

12. Import - Optionally import from another platform
13. Git - Initialize version control
14. Launch Dashboard - Open web UI

What gets created:

$SIGNET_WORKSPACE/
├── agent.yaml           # Configuration
├── AGENTS.md            # Agent identity
├── MEMORY.md            # Working memory
├── memory/
│   ├── memories.db      # SQLite database
│   └── scripts/         # Memory tools
├── harnesses/
├── hooks/               # OpenClaw hooks (if selected)
│   └── agent-memory/
└── .daemon/
    ├── plugins/         # Bundled core plugin registry
    └── logs/

If harnesses are selected, their configs are also created:

• Claude Code: ~/.claude/settings.json with hooks, ~/.claude/CLAUDE.md
• OpenCode: ~/.config/opencode/plugins/signet.mjs plugin, ~/.config/opencode/AGENTS.md
• OpenClaw: $SIGNET_WORKSPACE/hooks/agent-memory/ hook directory
• Codex: wrapper installed at ~/.config/signet/bin/codex with session hooks

signet configure

Interactive configuration editor for modifying $SIGNET_WORKSPACE/agent.yaml.

signet configure
signet config      # Alias

Sections:

1. Agent identity - Name and description
2. Harnesses - AI platform selection
3. Embedding provider - Ollama/OpenAI settings
4. Search settings - Alpha, top_k, min_score
5. Memory settings - Session budget, decay rate
6. View current config - Display agent.yaml contents

Changes are saved to agent.yaml immediately.

signet index <path>

Thin wrapper around graphiq index <path>. The command installs GraphIQ if it is missing, indexes the project into <path>/.graphiq/, enables the managed signet.graphiq plugin, and records that path as Signet’s active code project.

signet index ~/signet/signetai
signet index . --no-install

The GraphIQ index stays outside Signet memory and the main Signet database. Signet only stores plugin state and the active project pointer under $SIGNET_WORKSPACE/.daemon/graphiq/state.json.

signet graphiq

Manage the optional verified GraphIQ code retrieval plugin.

GraphIQ is maintained as a managed plugin by aaf2tbz, but remains optional and is not installed during setup unless the user opts in.

signet status

Show comprehensive status of the Signet installation.

signet status
signet status --path /custom/path

Options:

Output:

  ◈ signet v0.1.0
  own your agent. bring it anywhere.

  Status

  ● Daemon running
    PID: 12345
    Uptime: 2h 15m
    Dashboard: http://localhost:3850

  ✓ AGENTS.md
  ✓ agent.yaml
  ✓ memories.db

  Memories: 42
  Conversations: 7

  Path: /home/user/.agents

signet dashboard

Open the Signet web dashboard in your default browser.

signet dashboard
signet ui          # Alias

Options:

If the daemon is not running, it will be started automatically.

signet route

Inspect and control the shared inference router. Requires the daemon to be running for list, status, doctor, explain, and test.

signet route list
signet route status
signet route doctor
signet route explain "fix this bun test" --agent rose --task-class hard_coding
signet route test "summarize this transcript" --agent dot --task-class casual_chat --timeout 60000
signet route pin opus/opus46 --agent rose --task-class hard_coding
signet route unpin --agent rose --task-class hard_coding

Subcommands:

Common options:

Pins are stored under routing.agents.<agent>.pinnedTargets in $SIGNET_WORKSPACE/agent.yaml.

Daemon Commands

Daemon operations live under the signet daemon subcommand group. The top-level shortcuts still exist as backwards-compatible aliases, but the grouped form is the preferred surface.

signet daemon start
signet daemon stop
signet daemon restart
signet daemon status
signet daemon logs

# Backwards-compatible aliases
signet start
signet stop
signet restart
signet logs

signet daemon start

Start the Signet daemon if not already running.

  ◈ signet v0.1.0
  own your agent. bring it anywhere.

✔ Daemon started
  Dashboard: http://localhost:3850

Top-level alias: signet start

signet daemon stop

Stop the running Signet daemon.

Top-level alias: signet stop

signet daemon restart

Stop and start the daemon. Useful after installing an update.

Top-level alias: signet restart

signet daemon logs

View daemon logs.

signet daemon logs
signet daemon logs -n 100
signet daemon logs --follow
signet daemon logs --level warn
signet daemon logs --category memory

Top-level alias: signet logs

Options:

Service Installation

The daemon can be installed as a system service (systemd on Linux, launchd on macOS) using the daemon package’s bun scripts:

cd platform/daemon
bun run install:service    # Install as systemd/launchd service
bun run uninstall:service  # Remove the service

These are package-level scripts, not top-level signet CLI commands. They register a unit that starts the daemon automatically at login.

signet remember

Save a memory to the database. The daemon embeds it for vector search if an embedding provider is configured.

signet remember "User prefers dark mode"
signet remember "critical: never push to main" --critical
signet remember "deploy runs on Friday" --tags devops,deploy --who user

Options:

Output:

✔ Saved memory: mem_abc123 (embedded)
  Tags: devops,deploy

signet recall

Search memories using hybrid vector + keyword search.

signet recall "user preferences"
signet recall "release notes" --project /home/user/myapp --expand
signet recall "deploy process" --limit 5 --type decision
signet recall "auth" --tags backend --who claude-code --since 2026-01-01
signet recall "deploy checklist" --keyword-query "deploy OR rollback" --min-score 0.8
signet recall "secrets" --json

Primary controls:

Common refinements:

Advanced controls:

signet export / signet import

Export and import portable Signet bundles. This is the supported path for moving an agent between machines or backing up identity + memory state from the CLI.

signet export
signet export --json
signet import ./signet-export-2026-03-22
signet import ./signet-export-2026-03-22.json --json --conflict merge

signet export writes a portable bundle containing:

• identity files
• agent.yaml
• memories
• entities
• relations
• installed skills

signet import restores those files into $SIGNET_WORKSPACE/. Conflict handling for memories is controlled with --conflict:

• skip — keep existing memories and skip duplicates
• overwrite — replace matching memories
• merge — merge compatible records when supported

signet migrate-schema

Migrate an existing memory database to Signet’s unified schema. Useful when upgrading from an older version or copying $SIGNET_WORKSPACE/ between machines.

signet migrate-schema
signet migrate-schema --path /custom/path

Supported source schemas:

Migration is idempotent - safe to run multiple times. All existing memories are preserved. The daemon is stopped and restarted automatically during the process.

Output:

- Checking database schema...
  Migrating from python schema...
  ✓ Migrated 261 memories from python to core

  Migration complete!

signet migrate-vectors

Migrate existing BLOB-format embeddings to the sqlite-vec format. Run this once after upgrading from a version that stored vectors as raw BLOBs.

signet migrate-vectors
signet migrate-vectors --keep-blobs
signet migrate-vectors --dry-run

Options:

signet sync

Sync hooks, extensions, built-in template files, and skills to your $SIGNET_WORKSPACE/ directory, and re-register hooks for the active harnesses listed in agent.yaml. Run this after an upgrade if built-in skills appear stale or hooks need updating. If OpenClaw is still configured on the legacy Signet hook path, signet sync now migrates it to the plugin runtime path automatically so full lifecycle capture resumes. The command may report installed harnesses it detects on disk, but inactive harnesses are not modified.

signet sync

signet secret

Manage encrypted Secrets stored via the daemon, including 1Password service-account integration.

signet secret put OPENAI_API_KEY
signet secret put GITHUB_TOKEN ghp_...   # value inline
signet secret list
signet secret delete GITHUB_TOKEN
signet secret has OPENAI_API_KEY

# 1Password integration
signet secret onepassword connect
signet secret onepassword status
signet secret onepassword vaults
signet secret onepassword import --vault Engineering --prefix OP
signet secret onepassword disconnect

Subcommands:

A GITHUB_TOKEN secret is used by signet git to authenticate pushes to a remote repository.

signet skill

Manage agent Skills from the GitHub-based registry. Skills are installed to $SIGNET_WORKSPACE/skills/ and symlinked into harness config directories.

signet skill list
signet skill install browser-use
signet skill uninstall weather
signet skill search github
signet skill show <name>

Subcommands:

Registry search queries GitHub for repositories tagged agent-skill or containing a SKILL.md file. Unauthenticated searches are limited to 10 requests per minute.

signet git

Git sync management for the $SIGNET_WORKSPACE directory. A GITHUB_TOKEN secret must be set for push operations.

signet git status
signet git sync
signet git pull
signet git push
signet git enable
signet git enable --interval 600
signet git disable

Subcommands:

signet git enable options:

signet hook

Lifecycle hook commands for harness integration. These are called by connector packages automatically; you rarely need to invoke them directly.

signet hook session-start --harness claude-code
signet hook user-prompt-submit --harness claude-code
signet hook session-end --harness claude-code
signet hook pre-compaction --harness claude-code
signet hook compaction-complete --harness claude-code --summary "..."
signet hook synthesis
signet hook synthesis-complete --content "..."

Subcommands:

Most subcommands require -H, --harness <harness> identifying the calling platform (e.g. claude-code, opencode, openclaw). If the daemon is

When hook payloads are provided over stdin, the CLI now prefers canonical session_key / sessionKey fields before legacy session_id aliases. signet hook user-prompt-submit forwards preferred userMessage when it is provided, while still carrying legacy userPrompt compatibility fields. signet hook session-end forwards both stdin transcript_path / transcriptPath and inline transcript content for lossless capture. signet hook compaction-complete also forwards stdin cwd as the fallback project scope when transcript persistence has not landed yet. not running, hooks exit cleanly with code 0 so the harness is not blocked.

signet update

Check for updates, install them manually, or configure unattended auto-installs. Running signet update with no subcommand is equivalent to signet update check.

signet update               # same as check
signet update check
signet update check --force
signet update install
signet update status
signet update enable
signet update enable --interval 3600
signet update disable

Subcommands:

signet update check options:

signet update enable options:

After signet update install completes, a daemon restart is required to run the new version: signet daemon restart.

signet bypass

Toggle per-session hook bypass. When bypass is enabled for a session, all Signet hooks return empty no-op responses — the daemon is still running, but it stays silent for that session. MCP tools (memory_search, memory_store, etc.) continue to work normally.

signet bypass                   # List active sessions with bypass status
signet bypass --list            # Same as above
signet bypass <session-key>     # Enable bypass for a session
signet bypass --off <session-key>  # Disable bypass for a session

Subcommands:

You can also bypass hooks entirely at the process level using the SIGNET_BYPASS environment variable (see below).

signet embed

Manage memory embeddings. Requires the daemon to be running.

signet embed backfill
signet embed backfill --batch-size 100
signet embed backfill --dry-run
signet embed gaps

Subcommands:

signet embed backfill options:

After backfill completes, coverage is printed:

  Coverage: 100.0% (1200/1200 embedded)

Environment Variables

Exit Codes