nself model

Browse every available LLM — local Ollama models and cloud providers — and set the default for ɳClaw and the ai plugin.

Quick start

# List all available models (local + cloud)
nself model list

# Get details on a specific model
nself model info llama3.2

# Set the default model for the ai plugin and ɳClaw
nself model set llama3.2

# Interactively pick a model from a menu
nself model swap

Synopsis

nself model <SUBCOMMAND> [MODEL] [FLAGS]

Description

The ai plugin supports two model sources: Ollama (running locally alongside your stack) and cloud providers (OpenAI, Anthropic, Google, Groq, and others). nself model gives you a single interface to browse what is available, read model metadata, and choose the default that ɳClaw and your custom AI features use at runtime.

The active model is stored in NSELF_AI_MODEL in your .env file. Changing it via nself model set takes effect immediately for new requests — no rebuild required.

Cloud models require an API key in .env.secrets (e.g. OPENAI_API_KEY,ANTHROPIC_API_KEY). Local Ollama models require the Ollama service to be running, which ɳSelf manages automatically when ENABLE_OLLAMA=true.

Subcommands

nself model list

Show all models available to the current stack, grouped by source. Marks the active model with an asterisk.

nself model list
# LOCAL (Ollama)
#   llama3.2          3.8 GB   7B params   ← active (*)
#   llama3.2:70b      40 GB    70B params
#   mistral           3.8 GB   7B params
#   nomic-embed-text  274 MB   embedding
#
# CLOUD
#   openai/gpt-4o                requires: OPENAI_API_KEY
#   openai/gpt-4o-mini           requires: OPENAI_API_KEY
#   anthropic/claude-sonnet-4-6  requires: ANTHROPIC_API_KEY
#   anthropic/claude-haiku-4-5   requires: ANTHROPIC_API_KEY
#   google/gemini-1-5-pro        requires: GOOGLE_AI_KEY
#   groq/llama3-70b-8192         requires: GROQ_API_KEY

nself model info <MODEL>

Print detailed metadata for a model: parameter count, context window, supported capabilities, disk size (local), and pricing (cloud).

nself model info llama3.2
# Name:          llama3.2
# Source:        Ollama (local)
# Parameters:    7 billion
# Context:       128,000 tokens
# Capabilities:  chat, code, tool-use
# Disk:          3.8 GB
# Quantization:  Q4_K_M
# License:       Meta Llama 3.2

nself model info anthropic/claude-sonnet-4-6
# Name:          claude-sonnet-4-6
# Provider:      Anthropic
# Context:       200,000 tokens
# Capabilities:  chat, code, tool-use, vision
# Input price:   $3.00 / 1M tokens
# Output price:  $15.00 / 1M tokens
# Key required:  ANTHROPIC_API_KEY

nself model set <MODEL>

Set the default model. Writes NSELF_AI_MODEL=<model> to your active env file. The change takes effect immediately for new AI requests from ɳClaw and the ai plugin — no restart needed.

nself model set llama3.2
# ✓ NSELF_AI_MODEL=llama3.2 written to .env.dev

nself model set anthropic/claude-sonnet-4-6
# ✓ NSELF_AI_MODEL=anthropic/claude-sonnet-4-6 written to .env.dev
# ✓ ANTHROPIC_API_KEY detected in .env.secrets

nself model swap

Open an interactive menu to pick a model with arrow keys. Useful when you want to compare options before committing.

nself model swap
# ? Select the default AI model:
# ❯ llama3.2          (local, 3.8 GB)
#   llama3.2:70b      (local, 40 GB)
#   mistral           (local, 3.8 GB)
#   ─────────────────
#   openai/gpt-4o     (cloud, $5.00/1M in)
#   anthropic/...     (cloud, $3.00/1M in)

Adding a local model via Ollama

To use a model that is not yet downloaded, pull it with Ollama first, then switch to it:

# Pull a new model (Ollama must be running)
docker exec $(nself ps --name ollama) ollama pull phi3

# Verify it appears
nself model list --local

# Set it as default
nself model set phi3

Cloud model API keys

Place API keys in .env.secrets (gitignored). Never put them in .env.dev or .env.prod.

Environment variables

• NSELF_AI_MODEL — the active default model (set by nself model set)
• ENABLE_OLLAMA — set to true to start the Ollama service in the stack
• OLLAMA_HOST — Ollama API endpoint (default: http://ollama:11434)

Exit codes

• 0 — success
• 1 — model not found or API key missing
• 2 — invalid arguments
• 3 — ai plugin not installed (required for cloud models)

Related

• nself ai — run prompts and manage AI sessions
• nself plugin — install the ai plugin
• nself ai-studio — interactive AI testing UI