What it does
Hexis is a self-hosted, MCP-native control plane for the skills, tool manuals, context and permissions that AI agents use. Each of these is stored as a file in a git repository that the company owns, so git history serves as the audit trail for who changed what, when, and who approved it. Access is resolved per file, so an agent can only do what the person running it can do, and the agent never holds the credentials it uses. Anyone can propose a change, and agents can propose fixes too. On protected branches, the owners of the affected files review and approve a change before it ships.
The entry suits teams that want engineers and non-technical staff to browse and load plugins, suggest changes and manage access from one place, with any MCP-capable agent such as Claude or Cline connecting to it. The publisher contrasts this with MCP gateways, which let users consume tools but offer no built-in way to propose or review changes. A public demo instance, which can be signed into with a Google account, is available to try before deploying.
Running it requires Docker with Compose and an empty git repository on any host to hold the knowledge base; the app seeds it with a starter template on first run. Four required values must be filled into the environment file, including two generated secrets. For public HTTPS, the bundled Caddy proxy obtains certificates automatically, provided a DNS record points at the server and ports 80 and 443 are open. Alternatively, the app can sit behind an existing reverse proxy. The publisher also offers managed hosting.
Limitations and requirements
Requires Docker with Compose and an empty git repository on a host of your choice. Four required environment values must be set, including two generated secrets. The bundled HTTPS option needs a DNS record pointing at the server and ports 80 and 443 open to the internet. Behind your own reverse proxy, the origin values and proxy hop count must be configured. Cline connection requires OAuth sign-in in a browser. The public demo is shared and read-mostly, with visitors able to propose changes but not approve them.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
Hexis
Git-backed platform for skills, tools, and context for AI agents
v0.1.0As shown on registry.modelcontextprotocol.io when we last read it