An AI-assisted software development team built with Google ADK and Google Cloud.
Overview • Agents • Roadmap • Documentation
Multiagent Developer Bot coordinates specialized AI agents to assist with software planning, implementation guidance, security review, and operational debugging. The system exposes a consistent agent workflow across supported user platforms while keeping orchestration, persistence, and infrastructure concerns separate.
The Orchestrator is the primary user-facing agent. It delegates work to Frontend, Backend, DevOps, and Scrum specialists and combines their results into a coherent response.
Code produced by a specialist is reviewed by the Security Agent before it can be reported as compliant. The Security Agent is an independently deployed service that communicates with the Orchestrator through the A2A protocol. A failed review returns structured findings to the originating specialist for remediation. Reviews use a bounded retry workflow; unresolved findings are escalated to the user after the configured attempt limit. The Security Agent provides recommendations only and does not modify code or approve deployments.
The Cost Optimizer Agent is independently deployed with read-only access to approved billing exports, usage metrics, and resource metadata. It communicates with the Orchestrator through A2A and returns evidence-based savings estimates, tradeoffs, and recommendations. It cannot resize, stop, delete, or deploy cloud resources.
The Debug Agent is an independently deployed operational agent. It has read-only, least-privilege access to approved logs and produces evidence-based root-cause analysis with possible remediation steps. It remains isolated from the Orchestrator because it serves a different workflow and security boundary.
Status: Completed
- Initial multi-agent developer bot available through Discord
- Flask webhook deployed as a containerized AWS Lambda service
- Infrastructure managed with AWS Cloud Development Kit
- Direct OpenAI API integration for agent responses
- Established the initial Frontend, Backend, DevOps, and Scrum agent roles
Status: Work in progress
- FastAPI service deployed to Google Cloud Run
- Multi-agent orchestration with Google Agent Development Kit and Gemini
- Specialist agents for frontend, backend, DevOps, and Scrum
- Separate Security Agent for A2A-based security review
- Separate Cost Optimizer Agent for read-only cost analysis over A2A
- Separate Debug Agent for read-only operational analysis
- A2A support for independently deployed agent communication
- Cloud Firestore for project, task, workflow, and review state
- Secret Manager, Artifact Registry, Cloud Build, and Cloud Logging
- Discord support, with website and Slack adapters planned
Status: Planned
- Add Google Kubernetes Engine as a supported deployment target while preserving Cloud Run compatibility
- Run independently scalable agents and services with A2A communication
- Introduce Workload Identity, health probes, resource policies, autoscaling, network policies, and distributed observability
- Retain the v2 agent contracts and data model across deployment environments
Architecture, setup, and version-specific implementation notes are maintained in the project documentation.