From the Obot AI team
We've been building Obot as an MCP Gateway from the start. v0.22.0 is where we go bigger.
The problem we kept hearing from teams: the "client zoo." Developers spread across Cursor, Claude Desktop, Copilot, Windsurf — each with its own config, its own local MCP servers, its own skill definitions. No standardization. No visibility for admins. Skills getting duplicated and inconsistently authored across every team.
v0.22.0, which shipped last week, tackles that directly with three new capabilities:
→ Centrally managed skills — admins publish a curated catalog; users install to any local AI client with a single command
→ Fleet scanning — obot scan gives you a live inventory of every AI client, MCP server, and skill running across your org
→ Enterprise MCP controls — managed image pull secrets, external secret integration, OAuth inspector, user-defined headers
This is the beginning of a broader vision: a control plane for your entire AI tooling stack, not just MCP.
Top comments (0)