Gareth MorgansGovernance
AI agent governance for skills, prompts, MCPs and secrets
Most writing on AI governance covers policy frameworks, model risk and responsible-AI programmes. Those matter. But AI agent governance also needs control of the skills, prompts, MCPs, secrets and agents your people use every day, and that is where the gap usually sits.
Joyst closes that gap. It is an AI governance platform for those everyday tools, with permissions, folders and an audit trail. It is not a general GRC suite or a developer portal with scorecards.
Start free beta | AI governance
Related: ownership of skills and prompts | private MCP catalog vs marketplace | Joyst for platform teams.
An AI governance platform for skills, prompts and MCPs
Your team’s AI work lives in concrete things:
Prompts in a team library
Skills in a registry
MCP servers in a private catalogue
Credentials in a secrets vault
Packaged agents that pull those ingredients together
If those have no owners, no publishing step and no audit trail, governance stays a PDF. Real control means view, use and manage rights on the things people actually change.
Permissions and audit
Governance that works day to day includes:
Groups and folders, so large catalogues stay manageable
Draft and published states, so experiments stay out of everyday use
An audit trail of who changed which prompt, skill, MCP entry, secret reference or agent package
Approved catalogues, so people use only what the organisation has signed off
Read more about permissions and audit trail.
Vs model-risk GRC and vs IDP
Model-risk and responsible-AI programmes set standards. They rarely inventory your Cursor rules or which MCP server someone connected last week. For a practical starting point, see our AI governance framework.
Internal developer portals catalogue software services and scorecards. Joyst catalogues the AI tools your teams use. The two work together. See beside your IDP and AI registry vs developer portal. Joyst does not replace Backstage, Port, OpsLevel or Cortex.
What agentic AI governance means
Governing agents means governing what their packages include and who can publish or use them. That includes MCP tools and the way secrets are referenced. Joyst does not run multi-agent workflows. For the controls to put in place first, see our guide to securing AI agents.
The secrets vault is designed to keep keys out of the model’s context. Package setups with agents.
FAQ
What is an AI governance platform?
It is a product that controls permissions and audit for the AI tools your teams use: skills, prompts, MCPs, secrets and agents, held in a private catalogue for your organisation.
Is this responsible-AI ethics software?
No. Joyst does not sell ethics programmes or model-risk frameworks. It governs the skills, prompts, MCP servers, secrets and agents those programmes often leave uncontrolled.
Do we replace our GRC stack?
No. Joyst sits beside your policy and risk tools and controls the AI tools people use each day.