All configuration lives in the Settings app inside MeghaOS; there are no files to edit. Changes to the AI provider take effect immediately, without restarting.

​
AI provider

MeghaOS works with any major AI provider. In Settings → AI, pick a provider, paste a key (if the provider needs one), and choose a model: MeghaOS ships 25+ built-in providers — OpenRouter, OpenAI, Anthropic, Google Gemini, Groq, xAI/Grok, DeepSeek, Mistral, Together, Cerebras, NVIDIA, Perplexity, Cohere, and more. plus local runners (Ollama, LM Studio, vLLM) that need no key. Because most providers use the OpenAI-compatible format, you can also point MeghaOS at any compatible endpoint via the Custom OpenAI-compatible API option. See AI Providers for the full list.
The AI settings are live: switch provider or model in Settings and the next request uses it, no restart required.

​
Integrations

Every integration is optional. Add credentials in Settings → Integrations only for the capabilities you want to use.
Connect an email account (SMTP/IMAP) to send, read, and search mail. The built-in calendar works out of the box and can optionally sync with Google Calendar.
Add tokens for the chat platforms you use: Slack, Discord, Telegram, and Signal. With bot tokens configured, you can also message MeghaOS from those apps and get replies back.
Connect Twitter/X and LinkedIn to post updates and threads.
Add a news API key to power the live news ticker and news search.
Connect a GitHub account to manage repos, issues, and pull requests.
Some integrations are macOS-only because they use native Apple apps: iMessage, FaceTime, Apple Mail, Notes, Reminders, Contacts, and more. These work in the macOS edition without extra credentials. See Editions & Platforms.

​
Privacy & data

memory, and credentials, is stored locally on your device. Requests go only to the AI provider you configure (or stay entirely on-device with a local model).

​
Limits & behavior

A few system behaviors are tuned out of the box and don’t require configuration (things like how long a workflow step may run and how many steps run in parallel. These sensible defaults keep the assistant responsive.