# Hexis [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/Bevel-Software/Hexis  
**GitHub Stars:** 116  
**npm Downloads (last month):** 2642  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/hexis

## Description
Git-backed platform for skills, tools, and context for AI agents

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "hexis": {
    "command": "npx",
    "args": ["-y","@bevel-software/hexis-mcp"],
    "env": {
      "ADMIN_EMAIL": "",
      "ADMIN_PASSWORD": "",
      "JWT_SECRET": "",
      "SECRETS_ENC_KEY": ""
    }
  }
}
```

**Requires environment variables:** `ADMIN_EMAIL`, `ADMIN_PASSWORD`, `JWT_SECRET`, `SECRETS_ENC_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What Hexis MCP server does

Hexis MCP server provides a shared workspace for AI-agent skills, tool manuals, knowledge pages, permissions, and identity. The content is stored in a Git repository owned by the deploying organization, while Hexis provides the workspace and MCP access layer around it.

Agents can discover and load approved skills and company context that the connecting user is allowed to read. People can browse the same content, suggest new material, and manage access through roles or direct grants. This makes the platform suitable for teams that need a common source of agent instructions rather than separate prompt files maintained by each developer.

## How it works

Each skill, manual, permission-related definition, and knowledge item is represented as a file in a Git repository. Git history records changes, approvals, and reversions. On protected branches, a proposed change goes to the owners of the files it affects and becomes available only after approval.

The MCP connection exposes the workspace to an agent. Access is resolved for the person using the agent, so an agent receives only the skills, tools, and context that person can access. The README also states that the agent does not hold the credentials used for the underlying actions. Hexis can run on an organization’s infrastructure and can be placed behind its SSO.

## Setup and configuration

Self-hosting uses Docker Compose and an empty Git repository on a supported Git host, including GitHub, GitLab, Bitbucket, Azure DevOps, or a self-hosted option. The application seeds the repository with a starter template when it first starts.

A deployment requires values for `ADMIN_EMAIL`, `ADMIN_PASSWORD` when password login is enabled, `JWT_SECRET`, and `SECRETS_ENC_KEY`. The two secret values should be generated as random 32-byte base64 strings. Set `DOMAIN` when using the bundled proxy for a public HTTPS deployment. Operators can pin `HEXIS_VERSION` to avoid an unplanned upgrade when pulling newer images.

The repository also documents running the public demo, using Docker images published for releases, and building from source with a separate Compose configuration.

## Tools and capabilities

The documented MCP capabilities include:

- Discovering and loading accessible skills.
- Reading approved company context and knowledge.
- Making team skills and tool manuals available to connected agents.
- Proposing changes to skills or other managed content.
- Supporting human review and approval before changes reach protected branches.
- Applying role-based or direct access to shared workspace content.

The README does not provide a detailed list of individual MCP tool names or input schemas, so integrations should be checked against the deployment’s current MCP implementation.

## Limitations and notes

Hexis requires a Git repository for the knowledge base and Docker Compose for the documented deployment path. The public demo is shared and read-mostly: visitors can propose changes, while owners approve them. A production deployment still requires operators to manage its host, repository, secrets, backups, upgrades, and authentication choices.

Cline is documented as connecting through a remote Streamable HTTP MCP endpoint. The README also demonstrates connecting Claude over MCP, but client-specific setup details vary by deployment and client.

_Full upstream README: https://allmcps.com/mcp/hexis/readme_