LM Studio is a local runtime, not a service
LM Studio is a desktop application that downloads open-weight large language models and runs inference on your own hardware. There is no LM Studio-operated inference cloud, so the "data retention" framing that applies to OpenAI, Anthropic, or Google does not apply in the same shape.
LM Studio's privacy policy is explicit on this point: "None of your messages, chat histories, and documents are ever transmitted from your system - everything is saved locally on your device by default." The company also states that it cannot see any of your chats or documents within the app [1].
Once a model is downloaded, inference can run with no network connection. The model weights, the prompt, and the completion all stay on the machine.
No retention because there is no server
There is no LM Studio inference server in the loop, so there is no server-side prompt or completion store to configure or audit.
- Retention: None for inference content. Your prompts and completions are not sent to LM Studio and are not stored on any LM Studio server. Anything kept on disk (chat history, downloaded models) lives on your machine and is under your control.
- Training: LM Studio does not train models on your data. It serves open-weight models trained elsewhere by their respective publishers (Meta, Mistral AI, Qwen, and others).
- Zero data retention: Available by construction. Because inference is local, off-device transmission of your prompts and completions is zero, which is the strongest form of zero data retention available.
The network traffic LM Studio does generate - model search, downloads, and app update checks - contains none of your conversation content. LM Studio further notes that the application does not include telemetry or user-specific tracking, and that it does not build user profiles from your model preferences [1].
The local OpenAI-compatible server
LM Studio can run a local HTTP API server from its Developer tab, or via the lms server start CLI command. That server exposes OpenAI-compatible endpoints - Chat Completions, Completions, Embeddings, Responses, and more - at http://localhost:1234/v1 by default (the port is configurable in the app) [2][3].
Because the endpoint is OpenAI-compatible, any tool built for the OpenAI API can talk to a model running on your own machine simply by changing the base URL from OpenAI's servers to your local LM Studio instance [3]. No prompts cross the network boundary; the request goes to a process on the same device.
How Meetily uses LM Studio
Meetily's transcription path is local-by-default: audio is captured and transcribed entirely on your machine, and audio never leaves the device.
For summaries, Meetily supports bring-your-own-key (BYOK) against any OpenAI-compatible base URL. Point that base URL at LM Studio's local server (http://localhost:1234/v1) and the summarization step is served by a model running on your own hardware. With local transcription plus a local LM Studio summary endpoint, the entire pipeline - audio capture, transcription, summary - stays on the device.
This is the simplest answer to "does the summarization step have a retention policy I need to read?" The answer becomes "no, there is no remote service in the loop." For organizations subject to data-residency or processor-disclosure obligations, this path removes a class of compliance questions entirely. For everyone else, it is a low-friction way to get a private summary without sending transcript text to a third-party cloud API.
References
- "App Privacy Policy", LM Studio (Element Labs), https://lmstudio.ai/app-privacy (accessed 2026-06-29).
- "LM Studio as a Local LLM API Server", LM Studio Developer Docs, https://lmstudio.ai/docs/developer/core/server (accessed 2026-06-29).
- "OpenAI Compatibility Endpoints", LM Studio Developer Docs, https://lmstudio.ai/docs/developer/openai-compat (accessed 2026-06-29).