AI on privileged files.
Built for Mac and Windows.
Cloud AI assistants make you disclose privileged material to a third party. LMCP connects a local AI client to your Mail, Files and Notes on your device. Cloud AI models and optional connectors receive the content needed for authorized requests; desktop clients may also use cloud models.
Local-first · permission-aware · 188+ tools · macOS 13+ and Windows 10+
Why AI + privileged material is a confidentiality risk
Every mainstream way to use AI on legal work moves that work off your machine — and each hop is an outside disclosure you have to justify.
Cloud AI assistants send whatever you paste — including privileged material — to a third-party provider. Disclosing to an outside vendor can risk waiver and expands the confidentiality surface you have to defend.
Browser extensions & SaaS connectors route your mailbox and matter files through their servers. Another vendor with access to client confidences, another data-processing agreement to review.
Copy-paste into a chat window moves privileged text off your machine by hand — untracked, ungoverned, and hard to defend if opposing counsel asks where it went.
The local-first path
Keep the AI and the documents on the same Mac or Windows PC. On the local path there is no third-party cloud in the loop, so there is no outside disclosure to defend.
Runs on your computer
LMCP is a local MCP server for Mac and Windows. It works with supported apps and files on the device; cloud integrations are explicit and permissioned.
Choose where inference runs
A desktop AI client may use cloud models. Choose on-device inference and local integrations when the complete workflow must remain on-device.
Review connected services
Optional cloud AI providers and connectors receive the content required for authorized requests under their own terms. Review those agreements before using privileged material.
A straight answer on privilege
LMCP is architecture, not a compliance certificate, and nothing here is legal advice. No connector protects privilege by itself — that depends on your practices and jurisdiction. Local tools process data on your computer and return requested results to your selected AI client. Choose an on-device model and inspect all integrations for a fully local workflow. Use a cloud AI (ChatGPT, Claude.ai) and the material goes to that provider under their terms — LMCP is only the pipe.
Questions from GC and IT
Does using LMCP protect attorney-client privilege?
No software protects privilege on its own — that depends on your practices, jurisdiction, AI client and integrations. LMCP keeps local tools on your Mac or Windows PC and makes any cloud-connected path explicit so your team can review the boundary.
What if I use a web AI (ChatGPT, Claude.ai) instead of a desktop client?
Then the material you ask about is sent to that AI provider, and their terms apply — LMCP is only the pipe. For privilege-conscious work, choose an on-device AI model and inspect every integration for external transfers.
Where does the document text actually go?
On the local path, the MCP server works with data on your Mac or Windows PC and hands it to the AI client you selected. If you choose a web AI or cloud integration, that provider receives the data required for the request under its terms.
Can our firm review the data flow?
Yes — the local path is auditable: no outbound connection carries your document content when both inference and all selected tools operate on-device. Reach out and we’ll walk your IT/GC through exactly what the connector does and does not send.
Keep client confidences under your control.
Free for individual practitioners. Install in two minutes and pair it with a local AI model and local integrations when your workflow requires on-device processing. Desktop AI clients may also use cloud models.