Free
€0hosted, up to 3 peopleContext, skills, tools and access controls for Claude, ChatGPT, Cursor and any MCP agent, hosted by Bevel. Official source 1 ↗
Website-checked tool profile
Hexis
Open-source shared workspace for company knowledge, agent skills and tools, served to Claude, ChatGPT, Cursor and other agents over MCP.
Sources checked
- Pricing
- Free hosted plan for up to 3 people. Team costs €10 per person a month excluding VAT, billed only for active seats. Enterprise is custom-priced and adds SLAs, security-review help and rollout support. Self-hosting the open-source version is free. Official source 1 ↗
- Available as
- Hosted web app, plus a self-hosted option: deploy with Docker Compose on your own server, backed by a git repository on GitHub, GitLab, Bitbucket, Azure DevOps or a self-hosted git host. · Agents connect over MCP; any MCP-capable client works, including Claude Code, Codex, Cursor, Cline and ChatGPT. Official source 1 ↗ Official source 1 ↗
- Primary group
- Productivity
What it does
Hexis, from the Munich company Bevel, is a shared workspace that holds a company's knowledge pages, agent skills and tool connections, and serves them to AI agents such as Claude, ChatGPT, Cursor and Codex over a single MCP endpoint. Everything is stored as files in a git repository the company controls. Pages and skills have owners who approve changes, and folder-level permissions decide what each person and their agent can see. The core is open source under Apache-2.0 and can be self-hosted, or Bevel can host it.
Official source 1 ↗ · Official source 2 ↗Features
- Owner-approved knowledge pages. Company context lives in markdown pages that each have an owner. Edits go live only after the owner approves, and every version is kept. Official source 1 ↗
- Shared, reviewable skills. Teams keep agent skills in one place instead of on individual machines. Anyone, including an agent, can propose a change; the owners approve it before agents use it. There are 58 starter skills. Official source 1 ↗ · Official source 2 ↗
- One MCP endpoint for tools. Connect tools such as HubSpot, Notion, Jira, Linear, Google Drive, GitHub, Gmail and Zendesk, or any MCP server or API, once. Credentials stay on the server, and agents look tools up on demand instead of loading them all into context. Official source 1 ↗
- Fine-grained access. Rules per folder, file, skill and tool decide what each person can read or edit, and their agents inherit the same limits. Official source 1 ↗ · Official source 2 ↗
- Git storage and audit logs. Every page, skill and change is a commit in a repository you control, and audit logs show which agent used which tool and skill, with connections that can be revoked. Official source 1 ↗
- Open source and self-hostable. The Apache-2.0 core runs on your own infrastructure behind your SSO; the README walks through a Docker deployment. Official source 1 ↗ · Official source 2 ↗
Pricing & free limits
Team
€10per person / month, excl. VATEverything in Free; add or remove seats anytime and pay only for seats in use. Official source 1 ↗
Enterprise
Customcontact salesEverything in Team plus uptime and support SLAs, help with security and compliance reviews, and rollout support. Official source 1 ↗
Self-hosted
Freeopen source (Apache-2.0)Run all of Hexis on your own server with Docker; Bevel also offers to run a managed instance for you on request. Official source 1 ↗ · Official source 2 ↗
Billing details
Team pricing excludes VAT and is billed per active seat; seats can be added or removed at any time. Official source 1 ↗
Free access & limits
The hosted free plan covers up to 3 people. Official source 1 ↗
A public demo instance with a fictional company's data is open to anyone with a Google account; it is shared and read-mostly. Official source 1 ↗
Best for & limitations
Giving every employee's AI agent the same approved company context and skills, for example drafting a customer offer from a GTM skill, pricing guidance and past CRM deals. Official source 1 ↗
IT teams that need agent access governed: data in their own git server, per-folder permissions and audit logs. Official source 1 ↗
Letting non-technical staff propose and review skill changes without working in GitHub directly. Official source 1 ↗
- Needs MCP-capable agents. Hexis delivers context and tools over MCP, so it only helps with agents that can connect to an MCP server. Official source 1 ↗
- Self-hosting takes setup. Running it yourself needs a server with Docker, a git repository and secrets configuration; production use adds SSO, backups and health checks. Official source 1 ↗
- Thin public documentation. Beyond the product page and the GitHub README, there is little public documentation on the hosted plans. Official source 1 ↗
Similar tools
Tools like Hexis
Published tools that share its categories.
AncherAI knowledge base that saves articles, videos, posts and PDFs, answers from your own sources and drafts briefs and decks.
Visit 
BusabaseOpen-source database and review workspace where AI agents and people share data, docs and skills.
Visit 
WikiFixConfluence Cloud app that uses AI to find pages that contradict each other, duplicates and unowned pages, then fixes them with your approval.
Visit 
LUCILocal screen and meeting memory for AI agents, with desktop capture and searchable context.
Visit 
FaveNestA private saved-content library with on-device AI titles, summaries and tags on Apple devices.
Visit 
PipScreen-aware AI help for Mac, with spoken answers and visual guidance through app controls.
Visit