GCP Credit Request Agent Fleet

Cloud Credit Fleet models an enterprise Cloud FinOps and billing agent fleet rather than a collection of independent billing chatbots.

The project demonstrates a production-grade, multi-agent architecture designed to automate, evaluate, and govern Google Cloud Platform (GCP) promotional credits, billing adjustments, and credit program requests.

The project demonstrates:

Multi-agent coordination through a Root FinOps Agent.
Seven specialist-agent boundaries tailored for cloud financial operations.
Shared request and response contracts utilizing standardized JSON payloads.
Persistent case, ledger, and approval states.
Authority and approval gates (e.g., matching requested credit amounts against agent spending limits).
Audit, ledger, and escalation records to comply with financial governance.
Source-version validation (verifying credit program terms against current GCP policies).
Idempotent operations (preventing double-crediting of the same project or billing account).
Runtime configuration, logging, retry, and health controls.
Human escalation to FinOps leads or billing administrators for unresolved, high-value, or unsupported requests.

The credit adjustment workflow demonstrates characteristics of The Taskmaster track through autonomous routing across investigation, policy revalidation, billing adjustment calculations, approval, ledger correction, notification, and closure.

Inspiration

Cloud billing, promotional credits, and funding requests rarely belong to a single team, project owner, or automation script.

A single cloud credit request or billing discrepancy may require:

Project context (identifying the active resources and billing account owner).
Billing telemetry data (historical spend, SKU utilization, and credits already applied).
Credit policy interpretation (verifying if a startup or research grant program's criteria are met).
An adjustment calculation (estimating future run rates or calculating correct refund/credit values).
Human approval (when a credit request exceeds automatic approval boundaries).
Cloud Billing / KMS action (calling billing APIs safely or updating billing accounts).
Stakeholder notification (alerting the project owner and FinOps lead).
An immutable audit record (ensuring all adjustments are logged for financial compliance).

General-purpose LLM chatbots can explain credit rules, but an enterprise-ready Cloud Credit Request Agent must also determine:

Which specialist agent owns the specific credit evaluation.
Which terms of service or promotional program guidelines are currently authoritative.
Whether the project's historical billing data remains valid and unmanipulated.
Whether the agent has the financial delegation of authority (DoA) to automatically approve the request.
Which organizational approvals are required (e.g., $100 auto-approval vs. $5,000 director-level approval).
What happens when a Billing API or external database is temporarily unavailable.
Who owns the case when an edge case requires human intervention.
What ledger entries must be written to document the financial state change.

Cloud Credit Fleet was built to solve this operational gap. By using a Root Agent to coordinate specialized financial agents under a unified execution contract, the architecture prevents a single specialist from silently exceeding its authorized budget or treating an unverified adjustment as a successful transaction.