Today we’re releasing Linkly AI v0.2.0 — a version focused on connectivity.
What Problem This Solves
Linkly AI’s local knowledge base is exposed as an MCP service. In v0.1.x, this service was only accessible from your own machine — local AI tools like Cursor and Claude Code could connect directly, but cloud-based AI apps like ChatGPT and Claude.ai had no way to reach your local service.
v0.2.0 introduces the Remote Tunnel, solving this connectivity gap.
Three New Connection Modes
Remote Tunnel: ChatGPT / Claude.ai Accessing Your Local Knowledge Base
This is the core feature of v0.2.0.
The idea is straightforward: your desktop app establishes a persistent, encrypted WebSocket tunnel to mcp.linkly.ai. When ChatGPT or Claude.ai sends an MCP request to https://mcp.linkly.ai/mcp, the cloud service forwards it through the tunnel to your desktop, which executes the search and sends the result back.
Your documents always stay on your local machine. The tunnel only carries MCP tool call requests and responses — no document content passes through our servers.
Setup takes three steps:
- Go to linkly.ai, register, and generate an API Key from the Dashboard (format:
lkai_xxxx) - Open Linkly AI Desktop → Settings → MCP → paste the API Key → click “Connect Public Tunnel” → wait for the status to show ● Connected
- In ChatGPT or Claude.ai’s MCP settings, add:
https://mcp.linkly.ai/mcp. The first connection triggers an OAuth flow — log in once and you’re done
Full setup guide: Remote Access Documentation.
LAN Access: Share Your Knowledge Base Across Devices
If you only need to share your knowledge base across devices on your home network — no public internet needed — use LAN mode.
Once the desktop app enables the LAN service, other AI clients on the same Wi-Fi can directly access your knowledge base MCP service. No public tunnel, lower latency, and nothing passes through any server.
CLI Remote Mode: Command Line Goes Remote Too
The CLI tool now supports a --remote flag:
# Configure API Key once
linkly auth set-key lkai_xxxxxxxxxxxxxxxxxxxx
# Search through the public tunnel
linkly search "project proposal" --remote
# Read a document through the public tunnel
linkly read 42 --remoteThis lets you access your home machine’s knowledge base from anywhere — a server, another computer, anywhere with internet.
Typical Use Cases
Use case 1: Use your own documents in ChatGPT
You have a batch of industry research PDFs stored locally and want ChatGPT to answer questions based on them. With the tunnel active, ChatGPT directly calls your local search and read tools, retrieving from your private documents instead of relying on public internet information.
Use case 2: Access your desktop documents from your phone
You’re out and about, and you want to look up a contract clause on your desktop using Claude.ai on your phone. The desktop has the tunnel running, Claude.ai on your phone is already configured with MCP — just ask.
Use case 3: Team LAN sharing
A small team keeps their shared knowledge base on one machine. Everyone’s AI tools connect to that machine via LAN mode, sharing a single knowledge base.
Security
Your API Key is equivalent to your tunnel credential — do not share it with others. If it’s compromised, revoke it immediately in the Dashboard and generate a new one.
All tunnel traffic uses HTTPS/WSS encryption. You can disconnect the tunnel from the desktop app at any time. Once an API Key is revoked in the Dashboard, new connections and reconnects will be rejected.
mcp.linkly.ai is an MCP relay service officially maintained by Linkly AI. It only records call counts for display purposes and stores no document content.
Other Updates
v0.2.0 also includes bug fixes and stability improvements. See the full Changelog.
If you run into issues during setup, check the FAQ in the Remote Access Documentation.
From the Linkly AI team.