Do I need an account?+
Yes — the desktop app asks you to sign in once, and to pick a display name, so licensing and team features work. Nothing about your code goes with it: indexing, the 45 MCP tools, the local engine and your own API keys all run entirely on your machine.
Do I need an API key to use Tanvrit AI?+
No. Bringing your own key is a free-tier feature, not a paid one — add an Anthropic, OpenAI, Gemini, DeepSeek, Groq or Mistral key in Settings → Provider Keys and it works on Free. You can also run entirely on the bundled local runtime with no key at all. Pro buys priority routing on AI Mesh and higher limits, not access to your own keys.
What happens after the 30-day trial?+
Every account starts with a 30-day full-access trial. When it ends on the Free tier you see an upgrade prompt — and it is dismissible. Choose "Continue with my own API key" and local indexing, all 45 MCP tools, local models and your own provider keys keep working. Free users are never blocked.
Is SSO available on Team today?+
No, and we would rather say so than list it. Inbound enterprise SSO — consuming your identity provider — is a recorded non-goal until there is a paying requirement for it, so it is not built and not scheduled. Tanvrit does run its own OIDC provider, so a Tanvrit org account works as a sign-in; if you need us to consume your IdP instead, tell us and it becomes that paying requirement.
Is there a shared MCP server for a team?+
No. The MCP server binds to 127.0.0.1 on your own machine and that is deliberate — it is the reason your source never leaves it. What a team shares is knowledge, not a server: Team Knowledge sync shares curated decisions and context across the workspace, opt-in and off by default, while every index stays local.
Which models work?+
Locally, out of the box: the app ships its own inference runtime and pulls a coder model on first run, so there is no Ollama or Docker prerequisite — point it at an existing Ollama library only if you already have one. Remotely, on any tier including Free, with your own key: Claude, the GPT family, Gemini, DeepSeek, Groq and Mistral. Ten providers are registered in total, AI Mesh and the local runtimes included.
What hardware do I need?+
8 GB RAM is enough for small Ollama models. 16 GB is comfortable for qwen2.5-coder 7B. 32 GB+ unlocks the larger models. Apple Silicon and modern AMD/Intel CPUs both work; a discrete GPU helps but isn't required.
Is telemetry on?+
Off by default. If you opt in we collect anonymised counters (active platform, tool-use counts, error class) — never paths, URLs, emails, or code. The PII guard is fail-closed and rejects anything that looks like one. The endpoint is documented and can be blocked at the network layer.
Where does my data live?+
Your index, embeddings, conversation history and MCP credentials stay on your machine on every tier — the SQLDelight database is a local file. When you use a cloud provider, only the chunks your agent explicitly requests are sent, and only to the provider whose key you configured.
Why are the prices in rupees?+
Because that is the currency the checkout bills in, and it is what the in-app upgrade screen shows. Quoting a converted dollar figure here would drift away from the amount you are actually charged the moment the exchange rate moved.
Cancellation and refunds?+
Pro is month-to-month and cancels any time from Settings → Billing in the app, where the payment method and invoice history also live. Team is billed per active seat. Self-hosted licences are negotiated per deployment.
