Venice CLI

The Venice CLI is the official command-line interface for Venice. Chat with AI models, generate images, convert text to speech, transcribe audio, generate video, and more—all from your terminal, with optional end-to-end encryption.

GitHub: veniceai/venice-cli

Published as veniceai-cli on npm. MIT licensed.

Privacy-first

Optional E2EE and TEE attestation. No telemetry, no browser tracking.

Every modality

Chat, search, image, upscale, TTS, transcription, video, and embeddings

Scriptable

JSON and raw output, stdin piping, and shell completions for bash, zsh, fish

Quick start

Install the CLI

Install globally with npm:

npm install -g veniceai-cli

Or run it without installing:

npx veniceai-cli chat 'Hello, world!'

Get a Venice API key

Generate one from venice.ai/settings/api. See the API key guide for step-by-step instructions.

Configure your key

Run the interactive setup and paste your key when prompted, so it never appears in your shell history or process list:

venice config init

Or use an environment variable instead:

export VENICE_API_KEY=YOUR_API_KEY

Start chatting

venice chat "What is the meaning of life?"

Requires Node.js 18.0.0 or higher. VENICE_API_KEY overrides any value stored in the config file.

Commands

Command	What it does
`venice chat <prompt>`	Chat with any Venice text model, with tools, characters, and streaming
`venice search <query>`	Web search with AI synthesis and optional citations
`venice image <prompt>`	Generate an image from a text prompt
`venice upscale <file>`	Upscale an existing image up to 4×
`venice tts <text>`	Text-to-speech with 35+ voices
`venice transcribe <file>`	Speech-to-text with optional timestamps
`venice video <subcommand>`	Queue, poll, and download AI-generated video
`venice embeddings <text>`	Generate embedding vectors
`venice models`	List and filter the live model catalog
`venice tee <subcommand>`	Fetch and verify TEE attestation
`venice config <subcommand>`	Manage API key and defaults
`venice history <subcommand>`	List, show, export, or clear conversations
`venice usage`	Show token usage over time
`venice characters` / `venice voices`	List built-in personas and TTS voices
`venice completions <shell>`	Generate shell completions

Chat

# Basic chat
venice chat "Explain quantum computing in simple terms"

# Use a specific model and system prompt
venice chat -m deepseek-v3.2 -s "You are a helpful coding assistant" "Write fizzbuzz in Python"

# Continue the previous conversation
venice chat --continue "What about the next step?"

# Function calling with built-in tools
venice chat -t calculator,datetime "What's 25 * 4.5, and what day is it?"

# JSON output for scripting
venice chat -f json "List 3 colors" | jq '.content'

Common chat options:

Option	Description
`-m, --model <model>`	Model to use (default: `kimi-k2-5`)
`-s, --system <prompt>`	System prompt
`-c, --character <name>`	Character persona
`-t, --tools <tools>`	Comma-separated list of built-in tools
`--continue`	Continue the last conversation
`--web-search`	Enable web search for current information
`--no-stream`	Disable streaming output
`-f, --format <format>`	Output format (`pretty`, `json`, `markdown`, `raw`)
`--e2ee` / `--no-e2ee`	Force or disable end-to-end encryption
`--tee-verify`	Show TEE attestation details
`-q, --quiet`	Hide E2EE/TEE status messages

Web search

venice search "Latest developments in fusion energy"
venice search -n 10 --citations "Best practices for TypeScript"
venice search --scrape "Company research on Anthropic"

Image generation & upscaling

# Generate and save an image
venice image -o sunset.png "A serene mountain lake at sunset"

# Custom dimensions and model
venice image -w 1024 -h 768 -m flux-1-dev "Landscape photograph"

# Upscale an existing image 4×
venice upscale photo.jpg -s 4 -o photo_4x.jpg

Text-to-speech & transcription

# Speak with a specific voice
venice tts -v bf_emma -o greeting.mp3 "Good morning, everyone!"

# Pipe text in from stdin
echo "Text to speak" | venice tts -o output.mp3

# Transcribe audio with timestamps
venice transcribe -t recording.mp3

# Choose a model and language hint
venice transcribe -m openai/whisper-large-v3 -l es spanish_audio.mp3

Video generation

Video generation is asynchronous (queue-based).

# Queue a text-to-video generation
venice video generate -m veo3-fast-text-to-video "Cinematic sunset over mountains"

# Image-to-video with a reference image
venice video generate -m wan-2.6-image-to-video -i photo.jpg "The scene comes alive"

# Poll until done, then download
venice video status -w <queue_id>
venice video retrieve <queue_id> -o my_video.mp4

# List available video models
venice video models

Models & embeddings

# List and filter the catalog
venice models -t image
venice models --privacy
venice models --e2ee

# Generate embeddings
venice embeddings -o vectors.json "Text to embed"

Configuration

# Interactive setup
venice config init

# Set and inspect values
venice config set default_model kimi-k2-5
venice config get default_model
venice config show
venice config path

Key	Description
`api_key`	Your Venice API key
`default_model`	Default chat model
`default_image_model`	Default image generation model
`default_voice`	Default TTS voice
`output_format`	Default output format
`show_usage`	Show token usage after requests

Privacy: E2EE & TEE

Venice CLI is built for privacy. It ships no telemetry, stores your API key locally with restricted permissions, and supports Venice’s confidential compute models.

End-to-End Encryption (E2EE): Messages are encrypted client-side and decrypted only inside the TEE, so Venice cannot read your data. The CLI auto-enables E2EE when a model advertises the capability, generating an ephemeral key pair per session (ECDH + AES-GCM).
TEE Attestation: Cryptographically verify that a model runs in a secure enclave before sending data.

# Discover confidential-compute models
venice models --e2ee
venice models --tee

# Chat with E2EE (auto-enabled for capable models)
venice chat -m <e2ee-capable-model> "Your private message here"

# TEE-only mode: verify attestation without encryption
venice chat -m <e2ee-capable-model> --no-e2ee "TEE verified, not encrypted"

# Fetch and verify attestation directly
venice tee attestation <model>
venice tee verify <model>

E2EE mode disables tools and web search to preserve end-to-end encryption.

Output formats

Format	Use case
`pretty`	Colored, formatted output for interactive use (default)
`json`	Machine-readable JSON for scripting and piping
`markdown`	Markdown-formatted output for documentation
`raw`	Plain text with no decoration, for pipes

The CLI auto-detects when output is piped and switches to raw.

venice chat -f json "List items" | jq '.'
venice chat "Generate code" | pbcopy

Resources

GitHub

Source code, issues, and releases

npm

veniceai-cli

Venice MCP Server

Expose the same Venice tools to any MCP host

Venice API Reference

Full endpoint and parameter docs

Docs

Getting Started

Text & Chat

Image, Video & Audio

Search & RAG

Agents & Integrations

Coding Tools

Agent Tooling

SDKs & Frameworks

GitHub: veniceai/venice-cli

Privacy-first

Every modality

Scriptable

Quick start

Commands

Chat

Web search

Image generation & upscaling

Text-to-speech & transcription

Video generation

Models & embeddings

Configuration

Privacy: E2EE & TEE

Output formats

Resources

GitHub

npm

Venice MCP Server

Venice API Reference

GitHub: veniceai/venice-cli

Privacy-first

Every modality

Scriptable

​Quick start

​Commands

​Chat

​Web search

​Image generation & upscaling

​Text-to-speech & transcription

​Video generation

​Models & embeddings

​Configuration

​Privacy: E2EE & TEE

​Output formats

​Resources

GitHub

npm

Venice MCP Server

Venice API Reference

Quick start

Commands

Chat

Web search

Image generation & upscaling

Text-to-speech & transcription

Video generation

Models & embeddings

Configuration

Privacy: E2EE & TEE

Output formats

Resources