Platform Identifiers
Use these identifiers with the --platforms flag. All identifiers are case-insensitive.
| Identifier | Platform | URL | DEEP support |
|---|---|---|---|
| claude | Claude.ai | claude.ai | — |
| chatgpt | ChatGPT (OpenAI) | chatgpt.com | ✓ Deep Research |
| gemini | Google Gemini | gemini.google.com | ✓ Deep Research |
| grok | Grok (xAI) | grok.com | — |
| deepseek | DeepSeek | chat.deepseek.com | ✓ DeepThink |
| perplexity | Perplexity | perplexity.ai | — |
| copilot | Microsoft Copilot | copilot.microsoft.com | — |
CLI Flags
All flags can be passed when invoking /multai from Claude Code, or when running the orchestrator directly from the terminal.
| Flag | Type | Default | Description |
|---|---|---|---|
| --prompt | string | — | Inline prompt text to send to all platforms |
| --prompt-file | path | — | Path to a text file containing the prompt. Mutually exclusive with --prompt |
| --platforms | list | all 7 | Space-separated list of platform identifiers to run. Omit to run all 7 |
| --mode | enum | REGULAR | REGULAR or DEEP. DEEP activates Deep Research / DeepThink on supported platforms |
| --task-name | string | auto | Slug used in output filenames. Use alphanumeric characters and hyphens. Auto-derived from prompt if omitted |
| --output-dir | path | output/ | Directory where per-platform response files are written |
| --fresh | flag | false | Force a new browser state. Closes and reopens all platform tabs before the run |
| --with-fallback | flag | false | Enable browser-use agent fallback if Playwright automation fails for a platform. Requires API key in .env |
| --max-steps | int | 50 | Maximum steps for the browser-use agent during fallback. Increase for complex pages |
.env Configuration
Create a .env file in the MultAI project root to configure optional features. MultAI loads this file automatically on startup.
| Key | Required | Description |
|---|---|---|
| ANTHROPIC_API_KEY | Optional | Anthropic Claude API key. Enables Claude-powered browser-use agent fallback when Playwright fails |
| GOOGLE_API_KEY | Optional | Google Gemini API key. Enables Gemini-powered browser-use agent fallback |
| CDP_PORT | Optional | Chrome DevTools Protocol port. Default: 9222. Change if port is occupied |
| CDP_HOST | Optional | CDP host address. Default: 127.0.0.1. Do not change to 0.0.0.0 in production |
None of these keys are required for standard MultAI operation. The platform web UIs are accessed via your existing browser sessions — no API keys needed for Claude.ai, ChatGPT, Gemini, etc.
Output Files
MultAI writes to two directories relative to the project root:
| File / Path | Description |
|---|---|
| output/YYYY-MM-DD-task-platform.md | Per-platform response file. One file per platform per run. Contains the full response wrapped in <untrusted_platform_response> tags |
| archive/YYYY-MM-DD-task-archive.md | Collated archive containing all platform responses for a run. Primary artifact for review and consolidation |
| output/status.json | Current run status. Updated in real time as each platform completes. Contains per-platform status, error messages, and timing data |
| output/agent-fallback-log.json | Log of all browser-use agent fallback events. Includes platform, error reason, steps taken, and outcome |
Rate Limit State
MultAI persists rate limit tracking to disk so limits are respected across sessions:
| File / Path | Description |
|---|---|
| ~/.chrome-playwright/rate-limit-state.json | Per-platform rate limit state. Records hit timestamp and cooldown expiry for each platform. Updated automatically after every run |
| ~/.chrome-playwright/tab-state.json | Tracks the CDP tab IDs for each open platform tab. Used to reuse tabs between runs rather than opening new ones |
To reset all rate limits and force MultAI to treat all platforms as available:
rm ~/.chrome-playwright/rate-limit-state.json
To reset tab state (forces new tabs to be opened on next run):
rm ~/.chrome-playwright/tab-state.json
Or use the --fresh flag at runtime to reset browser state without deleting files manually.