OnlyMCP
A unified access layer product for MCP servers that hosts skills, agents, and memory on a remote server, allowing MCP clients such as Claude Code, Cursor, Codex, and ChatGPT to access them across devices through a single connection, and providing session and device management as well as authorization revocation.
Timeline (last 90 days)
OnlyMCP classifies content that needs to be synced across devices into three categories: skills, agents and memory.
OnlyMCP's new device authorization uses a one-time code confirmation model: when a new device logs in, a user-confirmed code is required.
OnlyMCP lists every session and every device and allows the user to revoke any one of them.
OnlyMCP's public information remains at the concept-demo level; the video does not explain where data is stored, whether it is end-to-end encrypted, or the specific integration method with each client, nor does it disclose pricing or open-source status.
The OAuth device authorization flow and the device management logic of mainstream password managers belong to the same category of practice as OnlyMCP's "one-time code confirmation + session revocation" design.
What OnlyMCP targets is the asset fragmentation problem that emerged after the spread of AI coding tools: skills, prompts, memory and MCP configurations are scattered across multiple devices and multiple clients, and migration costs rise as usage deepens.
All Facts (5)
OnlyMCP classifies content that needs to be synced across devices into three categories: skills, agents and memory.
50%UnverifiedOnlyMCP's new device authorization uses a one-time code confirmation model: when a new device logs in, a user-confirmed code is required.
50%UnverifiedOnlyMCP lists every session and every device and allows the user to revoke any one of them.
50%UnverifiedThe OAuth device authorization flow and the device management logic of mainstream password managers belong to the same category of practice as OnlyMCP's "one-time code confirmation + session revocation" design.
50%UnverifiedWhat OnlyMCP targets is the asset fragmentation problem that emerged after the spread of AI coding tools: skills, prompts, memory and MCP configurations are scattered across multiple devices and multiple clients, and migration costs rise as usage deepens.
50%