multisage

Query Multisage from your terminal. Get answers from a panel of AI experts (Claude, GPT, Gemini) with a single command.

Installation

npm install -g multisage

Or use without installing:

npx multisage "What's the best programming language for beginners?"

Setup

1. Create an API key at multisage.ai/settings
2. Set the environment variable:

export MULTISAGE_API_KEY=msk_your_key_here

Usage

# Ask a question
multisage "React vs Vue for a new project?"

# Get full expert details and stages
multisage --full "Should I use TypeScript?"

# Output as JSON (for scripting)
multisage --json "Best laptop for developers?"

# Full details as JSON
multisage --json --full "React vs Vue?"

# Pipe to a markdown renderer
multisage "Explain async/await" | glow

# Save to file
multisage "Best practices for REST APIs" > advice.md

# Use with a different server
multisage --api-url https://staging.multisage.ai "test question"

Options

| Flag | Short | Description | Default | |------|-------|-------------|---------| | --api-key <key> | -k | API key (overrides env var) | $MULTISAGE_API_KEY | | --api-url <url> | -u | Base URL for API | https://multisage.ai | | --stream | | Force streaming output (progressive rendering) | Auto-detect (TTY) | | --no-stream | | Force blocking output (wait for full response) | | | --full | -f | Include expert details and all stages | false | | --json | -j | Output as structured JSON (implies --no-stream) | false | | --quiet | -q | Suppress progress spinner | false | | --version | | Show version | | | --help | | Show help | |

Streaming Mode

By default, the CLI uses streaming mode when output is a terminal (TTY) and blocking mode when piped to a file or another command. Streaming mode shows content progressively as the AI panel works:

1. Quick Answer appears first, streaming character by character
2. Expert Responses (Claude, GPT, Gemini) appear as each expert completes
3. Synthesis streams the combined analysis
4. Debate streams if triggered

# Force streaming even when piped
multisage --stream "React vs Vue?" | tee output.md

# Force blocking (wait for complete response)
multisage --no-stream "React vs Vue?"

--json always uses blocking mode (streaming partial JSON is invalid).

Output Modes

Default (Markdown)

Prints the synthesized answer as markdown to stdout.

Full (--full)

Prints all stages: Quick Answer, Expert Responses (per expert), Synthesis, and Debate (if triggered). Works with both streaming and blocking modes.

JSON (--json)

Outputs the API response as JSON to stdout. Combine with --full for the complete response schema. Always uses blocking mode.

Error Codes

| Error | Message | |-------|---------| | Invalid API key | "Invalid API key. Check your key at: https://multisage.ai/settings" | | Insufficient credits | "Insufficient credits. Purchase more at: https://multisage.ai/settings" | | Rate limited | "Rate limit exceeded. Try again in N seconds." | | Concurrent limit | "Too many concurrent requests. Try again in a few seconds." | | Network error | "Could not connect to {url}." | | Timeout | "Request timed out after 300 seconds." |

All errors exit with code 1. With --json, errors are also output as JSON to stdout for programmatic handling.

Requirements

• Node.js 18 or later
• A Multisage API key (get one here)