convez.ai
AgenteIntegração & APIsAvançado
cmendezs

Payment Foundry

cmendezs/payment-foundry

Alinhamento multifuncional para integrações de pagamento, não apenas código

Ver no GitHub
InicialPluginScore · de 0 a 100

Agent de IA que guia equipes multifuncionais através de integrações com PSPs (Payment Service Providers) de forma estruturada. Captura requisitos de 8 stakeholders (Compliance, Fraud, Security, Finance, Backend, Frontend, Architecture, Head of Payments) e gera um Brief de Implementação executável com código real e reviews especializados.

Instalação

$npx convez-ai cmendezs/payment-foundry

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
42%
Último commit
há 2 meses
Versão
v0.3.0
Issues abertas
0
Contribuidores
1

Casos de uso

  • Integração de processadores de pagamento (Stripe, etc) com governança entre times
  • Captura estruturada de requisitos de compliance, fraud e segurança em projetos financeiros
  • Geração de documentação executável (Implementation Brief) com decisões rastreadas

Veredito

Ferramenta robusta para times que precisam integrar PSPs com rigor de governança. A abordagem de múltiplos stakeholders evita decisões técnicas isoladas, mas exige disciplina e disponibilidade de toda a equipe cross-funcional. Recomendado para empresas financeiras e fintechs com requisitos de compliance.

README

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

Payment Foundry

Cross-functional alignment, not just code.


What This Is

Most AI coding assistants generate a PSP integration straight from a prompt: a checkout form, a webhook handler, done. The code may run, but the decisions behind it (PCI scope, fraud rules, settlement currency, retry strategy) were never actually made by the people who own them.

Payment Foundry runs a structured engagement instead. A senior-level AI agent (the Engagement Manager) guides your technical team from scoping through implementation, and along the way captures requirements from the roles that have to live with the result: Compliance, Fraud, Security, Finance, Backend, Frontend, Architecture, and the Head of Payments.

One session. Eight perspectives. One Implementation Brief your engineers can execute from.


What You Get

CapabilityWhat it does
Structured engagement flowMoves your team through scoping, requirements, implementation, and review in a fixed, sensible order
Stakeholder requirements captureRecords each role's decisions and constraints to a dedicated file as the session progresses
PSP-grounded code examplesEvery code example is real and runnable, adapted from verified PSP reference content, never pseudocode
Specialist reviewsEight sub-agents respond approve, flag, or block, with reasoning, at the decision points that matter
Implementation BriefA single written deliverable covering decisions, code, open items, and anything unverified
Extensible to new PSPsAdd a new PSP by following the existing psps/<name>/ structure, no changes to core instructions needed

How a Session Flows

/start-session
      |
      v
  Scope & Constraints  ->  Stakeholder Requirements (one role at a time)
      |
      v
  /validate-context  ->  Verify PSP reference content against authoritative sources
      |
      v
  Core Payments -> Webhooks -> Platform / Terminal / Issuing (as in scope)
      |
      v
  Specialist Reviews (approve / flag / block)
      |
      v
/wrap-up  ->  Brief + Detailed Guide + Go-Live Checklist (outputs/<engagement>/)

Supported PSPs

PSPStatusProduct Lines
StripeAvailablePayments, Platform (Connect), Terminal, Issuing

More PSPs are planned. To request one or contribute, open an issue or pull request.


Prerequisites

  • One of the supported AI coding agents: Claude Code, Google Antigravity, AWS Kiro, or Mistral Vibe. See setup/other-agents.md for per-tool setup notes
  • A supported PSP (v1: Stripe only)
  • Stripe test API keys (publishable + secret)

Getting Started (Four Steps)

1. Clone and enter the directory

git clone https://github.com/cmendezs/payment-foundry.git
cd payment-foundry

2. Set up your environment

cp .env.example .env
# Open .env and fill in your Stripe test keys
# See setup/environment-keys.md for details

3. Read the first-session checklist

Open setup/first-session-checklist.md and work through it before your first session. It takes about ten minutes and prevents the most common setup issues.

4. Launch and start your engagement

claude

Then in the Claude Code session:

/start-session

Using Google Antigravity, AWS Kiro, or Mistral Vibe instead? The same /start-session, /validate-context, and /wrap-up flow is available in each, see setup/other-agents.md for the equivalent launch steps and skill locations.


Slash Commands

Three commands cover the full engagement lifecycle.

CommandWhen to runWhat it does
/start-sessionBeginning of every engagementIdentifies the PSP, scopes the engagement, captures stakeholder requirements one role at a time, and proposes the implementation sequence
/validate-contextAfter /start-session, before implementation beginsChecks the PSP-specific facts in scope (status, pricing, capability gating, header strings, API versions) against the PSP's authoritative sources, and records what is verified, unverified, or blocked
/wrap-upEnd of the engagementCollects open items, documents sub-agent outcomes, and produces three artifacts under outputs/<engagement>/: an executive Implementation Brief, a code-heavy Detailed Implementation Guide, and a Go-Live Readiness Checklist

Everything between /start-session and /wrap-up is handled conversationally by the Engagement Manager: implementation guidance, code examples, and specialist reviews as decisions arise.


How a Session Works

A typical engagement moves through these stages in order:

  1. Scope - Use case, tech stack, markets, currencies, timeline, team size
  2. Stakeholder requirements - Head of Payments, Compliance, Fraud, Backend, Frontend, Architecture, Security, Finance captured conversationally and saved as reference files
  3. Context validation - /validate-context checks the PSP-specific facts in scope against authoritative sources and records verified, unverified, and blocked items
  4. Core payments - Payment Intents, checkout UI, confirmation handling
  5. Webhooks - Event handling, order state reconciliation, retries
  6. Platform flows (if applicable) - Connect, multi-party payouts
  7. In-person flows (if applicable) - Terminal, reader management
  8. Card issuing (if applicable) - Issued cards, spend controls, authorizations
  9. Specialist reviews - Each sub-agent loads its requirements file, reviews the relevant decisions, and outputs approve / flag / block with reasoning
  10. Engagement Artifacts - Executive Implementation Brief, code-heavy Detailed Implementation Guide, and per-engagement Go-Live Readiness Checklist

The Engagement Manager proposes this sequence at the start and adapts it to what is actually in scope for your team.


Specialist Sub-Agents

Eight specialists are available for cross-functional review. The Engagement Manager invokes them at the right decision points: you do not need to call them directly.

SpecialistReviews
Head of PaymentsKPI monitoring, migration risk, operational governance
Compliance OfficerPCI scope, SCA/3DS2, data residency, audit trail
Fraud OfficerRisk rules, 3DS strategy, dispute and chargeback process
Security OfficerSecret management, webhook validation, fraud controls
Solution ArchitectIntegration patterns, failure modes, scalability
Frontend DeveloperCheckout UX, error handling, accessibility, localization
Backend DeveloperIdempotency, webhook processing, retries, reconciliation
Finance and TreasurySettlement, multi-currency, payouts, reporting, tax

Each specialist produces an outcome: approve, flag (proceed with conditions), or block (stop until resolved). The Engagement Manager helps you resolve flags and blocks before moving on.


Directory Structure

payment-foundry/
├── README.md                        # You are here
├── CLAUDE.md                        # Engagement Manager instructions
├── AGENTS.md                        # Shared agent instructions pointer, read by Mistral Vibe, AWS Kiro, and other AGENTS.md-aware tools
├── .env.example                     # Copy to .env and fill in your keys
│
├── setup/                           # Run once before your first session
│   ├── installation-guide.md
│   ├── first-session-checklist.md
│   ├── environment-keys.md
│   └── other-agents.md              # Per-tool notes: Claude Code, Antigravity, Kiro, Vibe
│
├── skills/
│   └── payment-foundry/             # Source of truth for the three skills
│       ├── start-session/SKILL.md      # /start-session command
│       ├── validate-context/SKILL.md   # /validate-context command
│       └── wrap-up/SKILL.md            # /wrap-up command
│
├── scripts/
│   └── setup-agents.sh              # Distributes skills/payment-foundry/ to each tool below
│
├── .claude/skills/                  # Claude Code copy (generated by scripts/setup-agents.sh)
├── .agents/skills/payment-foundry/  # Google Antigravity / AWS Kiro copy (generated by scripts/setup-agents.sh)
├── .vibe/agents/payment-foundry.toml # Mistral Vibe subagent profile (generated by scripts/setup-agents.sh)
│
├── sub-agents/                      # Specialist reviewer definitions
│   ├── README.md                    # Invocation procedure
│   ├── head-of-payments.md
│   ├── compliance-officer.md
│   ├── fraud-officer.md
│   ├── security-officer.md
│   ├── solution-architect.md
│   ├── frontend-developer.md
│   ├── backend-developer.md
│   └── finance-treasury.md
│
├── psps/                            # PSP reference content, loaded at runtime
│   └── stripe/
│       ├── README.md                # Index: which file covers what
│       ├── payments.md
│       ├── platform.md
│       ├── terminal.md
│       └── issuing.md
│
├── context/                         # Scoping and requirements templates
│   ├── business-info.md              # /start-session scoping guide
│   ├── go-live-checklist-template.md # Source template for go-live checklist
│   ├── head-of-payments-requirements.md
│   ├── compliance-officer-requirements.md
│   ├── fraud-officer-requirements.md
│   ├── backend-developer-requirements.md
│   ├── frontend-developer-requirements.md
│   ├── solution-architect-requirements.md
│   ├── security-officer-requirements.md
│   └── finance-treasury-requirements.md
│
└── outputs/
    ├── <engagement>-*-requirements.md      # Captured per session, per role
    ├── <engagement>-context-validation.md  # Produced by /validate-context
    └── <engagement>/                       # Per-engagement folder, produced by /wrap-up
        ├── implementation-brief.md         # Executive layer
        ├── implementation-detailed.md      # Developer manual with code
        └── go-live-checklist.md            # Adapted from the template

Company information lives in context/business-info.md and is updated in place across engagements, never copied per engagement.


Adding a New PSP

Create a folder under psps/<name>/ with a README.md index and one file per product line. Follow the same structure as psps/stripe/. No changes are needed to CLAUDE.md or sub-agents/.


License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

Ver o original no GitHub

Bom para

Relacionados