Concepts
AI Agent Governance
botYguard is a security and governance layer for AI agents. It sits between hosted or MCP-connected agents and sensitive operations so teams can decide what agents may access before tool calls, provider keys, or connected systems are used.
Use this page when you need to explain how botYguard helps an organization adopt agentic workflows without handing agents unrestricted access to secrets, data, and tools.
Key Concepts
- Agent profile: describes the agent type and how botYguard should govern it.
- Policy: defines which resources, integrations, and actions are allowed, denied, audited, or sent for approval.
- Approval: pauses sensitive work until a human or system approves the action.
- Secret handling: keeps provider keys and integration credentials out of prompts and external agent instructions.
- Audit visibility: records what the agent requested, what policy decided, who approved it, and what ran.
Why Governance Matters
AI agents can call tools, read workspace data, use credentials, and take multi-step actions. Local prompt rules are not enough when an agent has real access to SaaS systems, model providers, internal APIs, or private context.
botYguard moves governance to a dedicated control layer. The agent can still complete useful work, but sensitive access flows through policy, approvals, privacy controls, and audit logging.
Common Governance Flow
- A user asks an agent to complete a task.
- The hosted or MCP-connected agent requests access to a tool or integration.
- botYguard checks profile, policy, secret, and approval requirements.
- Allowed work continues with the approved scope.
- Sensitive work is denied, audited, or routed for approval.
- botYguard records the decision and execution evidence.