تخطَّ إلى المحتوى

nself claw

هذا المحتوى غير متوفر بلغتك بعد.

Manage ɳClaw AI assistant.

nself claw <subcommand> [flags] [args]

nself claw is the operator and end-user control plane for ɳClaw, the self-hosted AI assistant. It covers pairing client apps to a server, enabling the web UI for first-time setup, sending one-shot prompts, opening interactive chats, browsing topics and memories, managing API keys, exposing an OpenAI-compatible local proxy, running an MCP server for tool integrations, and exporting all data.

The CLI talks to the claw plugin over its HTTP API. Authentication is by API key (env nself or ~/.nself/claw/config.yaml). For first-time setup, run �P1� claw enable on the server to open a 10-minute window for account creation, then �P2� claw pair --qr to connect a mobile client.

nself claw mcp exposes ɳClaw memory, topics, and chat as Model Context Protocol tools so any MCP-aware client (Claude Code, IDEs) can read and write ɳClaw context. �P1� claw proxy exposes an OpenAI-compatible local endpoint so any client expecting OpenAI can route through ɳClaw without code changes.

NameDescription
pairGenerate a pairing code for ɳClaw clients
enableTemporarily enable the web UI for first-time account setup
prompt [question]Send a single prompt and print the response
chatStart an interactive chat session
configShow or modify ɳClaw CLI configuration
config set <key> <value>Set a configuration value (api-key, server)
topicsList topics
topics search <query>Search topics by name
memoriesList memories
memories search <query>Search memories by content
keysList API keys
keys createCreate a new API key
keys revoke <id>Revoke an API key
statusShow ɳClaw server status and health
proxy [port]Start a local OpenAI-compatible proxy
mcpStart an MCP server for ɳClaw
exportExport all ɳClaw data
FlagDefaultDescription
--qrfalseDisplay QR code in terminal
--directfalseSkip cloud relay, local pairing only
FlagDefaultDescription
--minutes10Minutes the enable lasts (1-60)
FlagDefaultDescription
--jsonfalseOutput full JSON response
--model""Override model name
--topic""Topic context
--no-memoryfalseSkip memory injection
--system""Custom system prompt
--rawfalseNo formatting, raw text only
--timeout120sRequest timeout
--file, -f""Read prompt from file
--no-streamfalseWait for complete response
FlagDefaultDescription
--topic""Start in specific topic
--model""Use specific model
--resumefalseResume last conversation
--session""Resume specific session ID
FlagDefaultDescription
--name""Name for the API key (required)
FlagDefaultDescription
--transportstdioTransport mode: stdio or sse
--port8899Port for SSE transport
FlagDefaultDescription
--formatjsonOutput format: json or csv

claw config, claw config set, claw topics, claw topics search, claw memories, claw memories search, claw keys, claw keys revoke, claw status, and claw proxy accept no flags beyond positional arguments.

Terminal window
# First-time server setup: open the web UI for 10 minutes to create an account
nself claw unlock
# Pair a mobile or desktop client (prints QR code)
nself claw pair
# Save API key and server URL on a workstation
nself claw config set api-key nself_claw_xxxxxxxxxxxxxxxx
nself claw config set server https://claw.example.com
# Send a one-shot prompt and stream the answer
nself claw prompt "Summarize the project README"
# Pipe stdin through nClaw with a custom system prompt
cat notes.txt | nself claw prompt "Extract decisions" --system "You are a precise note-taker"
# Open the interactive chat REPL
nself claw chat
# Search memories
nself claw memories search "Postgres tuning"
# Run as an MCP server for Claude Code
nself claw mcp --transport stdio
# Run an OpenAI-compatible proxy on port 9000
nself claw proxy 9000
# Back up all nClaw data
nself claw export > claw-backup.json

Commands | Home