Cost And Token Governance

MeshGuard attributes model spend to agents, teams, business units, projects, and customers.

Provider Coverage

• OpenAI and Azure OpenAI.
• Anthropic.
• AWS Bedrock.
• GCP Vertex AI.
• Cohere.
• Mistral.
• Local or self-hosted models through token estimation.

Controls

• Daily, weekly, monthly, and rolling-window budgets.
• Soft threshold alerts.
• Hard caps enforced as policy denials.
• Spend anomaly detection.
• Chargeback exports for finance.

Policy Example

apiVersion: governance.toolkit/v1
kind: Policy
metadata:
  name: monthly-token-budget
spec:
  defaultEffect: deny
  rules:
    - effect: deny
      actions:
        - model:invoke
      when:
        meshguard.app/v1:
          budgetRemainingUsd:
            lte: 0
    - effect: allow
      actions:
        - model:invoke