convez.ai
MCP ServerDevOps & InfraestruturaAvançado
shubhamjaggi

Gentepede MCP

shubhamjaggi/gentepede-mcp

Gera infraestrutura AWS segura e validada a partir do stack da aplicação

Ver no GitHub
BomPluginScore · de 0 a 100

Servidor MCP em Kotlin que mapeia tech stacks de aplicações (Spring Boot, Ktor, Node.js, FastAPI) para infraestrutura AWS totalmente provisionada via Terraform e Helm. Inclui pipeline completo de validação, segurança (checkov/kube-score), estimativa de custos (Infracost) e detecção de drift.

Instalação

$npx convez-ai shubhamjaggi/gentepede-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
100%
Último commit
há 2 semanas
Versão
—
Issues abertas
8
Contribuidores
3

Casos de uso

  • Provisionar infraestrutura AWS automatizada para diferentes tech stacks
  • Validar segurança de configurações Terraform com checkov antes de aplicar
  • Estimar custos em tempo real e detectar drift em deployments EKS
  • Gerar charts Helm production-ready para Kubernetes com validação kube-score

Veredito

Ferramenta robusta e bem-estruturada para times DevOps que precisam automatizar provisionamento AWS com guardrails de segurança e custo. Requer profundo conhecimento de Terraform, Kubernetes e AWS, mas reduz drasticamente overhead de IaC manual.

README

A tradução para o português deste README ainda está na fila.

Gentepede MCP

CI Lint Blueprint Verification License: MIT Java 21 Kotlin

A Kotlin-based MCP server that acts as a deterministic Knowledge Engine for enterprise cloud architectures. Maps application tech-stacks (Spring Boot, Ktor, Node.js, FastAPI) to fully provisioned, secure AWS infrastructure via Terraform — and for EKS blueprints also generates a production-ready Helm chart alongside the Terraform. Features a full plan-review-apply pipeline, live cost estimation via Infracost, drift detection, and credential transparency.


Architecture Diagram

MCP Client (Claude Desktop) → Gentepede Server → Blueprint Engine
                                                         │
                                      ┌──────────────────┴──────────────────┐
                                 Terraform                               Helm Chart
                                      │                                      │
                           ┌──────────┴──────────┐                kube-score lint
                      checkov lint          infracost                        │
                                      │                           helm upgrade --install
                               plan → apply                                  │
                                      │                              EKS (real AWS)
                                   real AWS

Prerequisites

Install all of the following before building:

ToolPurposeInstall
Java 21JVM runtime for the serverSDKMAN / Homebrew / winget
TerraformIaC provisioningbrew install terraform
checkovSecurity linterpip install checkov
HelmKubernetes packagingbrew install helm
kube-scoreK8s manifest linterGitHub releases
infracostCost estimationbrew install infracost
kubectlKubernetes CLIbrew install kubectl
helm-diff (optional)K8s drift detectionhelm plugin install https://github.com/databus23/helm-diff

Quick Start

# 1. Clone and build
git clone https://github.com/shubhamjaggi/gentepede-mcp
cd gentepede-mcp
./gradlew shadowJar

# 2. Configure Claude Desktop
# Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
# {
#   "mcpServers": {
#     "gentepede": {
#       "command": "java",
#       "args": ["-jar", "/path/to/build/libs/gentepede-mcp-all.jar"],
#       "env": { "AWS_PROFILE": "my-profile", "AWS_DEFAULT_REGION": "us-east-1" }
#     }
#   }
# }

# 3. Restart Claude Desktop, then ask:
# "List available Gentepede blueprints"
# "Generate a Spring Boot + Postgres ECS infrastructure called 'my-api'"

See docs/11-end-to-end-walkthrough.md for the complete guide.


Deployment Workflow

generate_infrastructure_package   → Creates Terraform + Helm files in workspace
        ↓
validate_infrastructure_package   → terraform validate + checkov + kube-score (no AWS calls)
        ↓
plan_infrastructure_package       → terraform plan + infracost cost estimate
        ↓  (review the output)
apply_infrastructure_package      → terraform apply + helm upgrade --install
        ↓
detect_drift                      → Ongoing: check for manual AWS changes
        ↓
destroy_infrastructure_package    → terraform destroy + helm uninstall

audit_infrastructure_package      → Standalone security report (any time)

Configuration Matrix

Environment VariableDefaultDescription
AWS_DEFAULT_REGIONus-east-1AWS region for deployments
AWS_PROFILEdefaultAWS CLI named profile
TF_LOG(unset)Set to DEBUG to see Terraform API calls
KUBECONFIG~/.kube/configKubernetes config file location

Supported Blueprints

Blueprint IDFrameworkOutputAWS ResourcesProviderLast VerifiedEst. Monthly Cost
springboot-postgresSpring BootTerraform onlyECS Fargate, RDS PostgreSQL, ALB, VPC, KMS5.82.02026-06~$70-120
ktor-dynamodbKtorTerraform onlyECS Fargate, DynamoDB, ALB, VPC, KMS5.82.02026-06~$40-80
nodejs-s3Node.jsTerraform onlyLambda, API Gateway, S3, CloudFront, KMS5.82.02026-06~$5-30
fastapi-redisFastAPITerraform onlyECS Fargate, ElastiCache Redis, ALB, VPC, KMS5.82.02026-06~$50-90
springboot-eksSpring BootTerraform + HelmEKS Cluster, RDS PostgreSQL, ALB Ingress, VPC5.82.02026-06~$150-250
nodejs-eksNode.jsTerraform + HelmEKS Cluster, S3, CloudFront, ALB Ingress, VPC5.82.02026-06~$100-200

Security Model

Gentepede embeds security best practices directly in every Terraform template and Helm chart — you cannot generate insecure infrastructure without modifying the templates. checkov runs as a gate before planning (blocks on HIGH/CRITICAL); kube-score validates Kubernetes manifests. AWS credential identity is confirmed before every operation that contacts AWS, and plan file checksums prevent stale plan application. See docs/08-security-model.md.


Documentation

FileDescription
docs/00-glossary.mdPlain-English definitions of all technical terms
docs/01-what-is-this.mdProblem, solution, example interaction, who it's for
docs/02-architecture.mdBeginner diagram + technical call-graph + data flow
docs/03-how-mcp-works.mdMCP protocol, stdio transport, Claude Desktop config
docs/04-blueprints-guide.mdBlueprint schema field-by-field, outputType comparison
docs/05-terraform-guide.mdResource-by-resource walkthrough, state management
docs/06-kubernetes-guide.mdECS vs EKS, Helm chart walkthrough, Kubernetes resources
docs/07-tools-reference.mdAll 8 tools: full input/output, workflow diagram
docs/08-security-model.mdcheckov + kube-score tables, credential pre-flight
docs/09-adding-blueprints.mdWorked example: adding go-dynamodb end-to-end
docs/10-troubleshooting.mdEvery common error with exact message, cause, fix
docs/11-end-to-end-walkthrough.mdComplete Phase 1-6 walkthrough from fresh machine to AWS deployment
docs/12-development-guide.mdBuild, test, debug, project structure for contributors
docs/13-faq.mdCommon questions from users and contributors
docs/14-blueprint-to-resource-map.mdFull mapping: which blueprint provisions which AWS services and why
docs/15-tool-architecture.mdEnd-to-end architecture of all 8 tools: every layer, every CLI call, every file
docs/16-contributor-sync-guide.mdComplete sync checklist for every contributor change type (blueprint, template family, tool, Helm chart, provider bump)
docs/17-github-actions-guide.mdPlain-English explanation of every CI workflow: what it does, when it runs, and why

Contributing

Contributions welcome — especially new blueprints for additional tech stacks. Start with CONTRIBUTING.md, and see docs/09-adding-blueprints.md for the step-by-step blueprint guide and PR checklist.

The architecture deliberately separates concerns: Engine.kt (thin MCP handler), InfrastructureService.kt (business logic, fully testable without MCP server), Validator.kt (CLI output parsing), and Models.kt (shared types). Keep this separation when contributing.

This project follows a Code of Conduct. To report a security vulnerability, see SECURITY.md — please do not open a public issue.


License

MIT License. See LICENSE for details.

Ver o original no GitHub

Bom para

Relacionados