Infrastructure as Code just got an upgrade to Infrastructure as Agents. Project Sentinel is an autonomous, multi-agent "Crew" designed to solve the two biggest headaches in cloud engineering: Manual Toil and Budget Drift. By orchestrating Google Gemini, Claude 3.5, and E2B sandboxed runtimes, this system allows users to provision complex GCP resources using natural language while maintaining 100% budget compliance—autonomously.
graph TD
%% Input and Analysis Phase
A --> B{Architect: Gemini}
B -- Ingests Requirements & Log Data --> C
B -- Defines Data Storage Strategy --> GCS_Logs
%% Generation Phase
C --> D
D -- Generates HCL: IAM, Svc Acct, GKE, Namespaces --> E
%% Validation & Version Control
subgraph Execution_Sandbox
E -- Runs 'Terraform Init & Plan' --> F{Validation?}
F -- Errors Found --> D
F -- Success --> GH
end
%% Provisioning Phase
GH -- Push Code & Open PR --> GPR
GPR -- Human Approval Gate --> TA
%% GCP Resource Layer
subgraph Google_Cloud_Platform [Google Cloud Platform]
TA --> IAM
IAM --> GKE[GKE Cluster & Namespaces]
TA --> ST
end
%% CI/CD and Monitoring Layer
GKE --> CF[Codefresh Pipeline]
CF -- Blue-Green Deployment --> PROD
PROD --> MON{Monitoring Agent}
MON -- Fetches Status API --> GRAF
GRAF -- Alert/Health Output --> A