serve7-mcp-connectors
Serve7 MCP Connector Library
Herramientas reutilizables de MCP para Amazon Connect AI Agents y Bedrock AgentCore. Siete conectores genéricos —consulta de cliente, creación de un ticket de soporte, comprobación del estado del ticket, programación de una cita, actualización de un registro de CRM, recuperación de conocimiento y creación de una tarea—, construidos aquí una sola vez y reutilizados en todas las interacciones con los clientes de Serve7, en lugar de tener que conectarlos desde cero en cada proyecto.
Este es el proyecto n.º 6 de la lista de ideas de prácticas de otoño de 2026 («Serve7 MCP Connector Library»). Es una primera versión funcional: cada conector funciona hoy de extremo a extremo contra un backend simulado en memoria (sin cuenta de AWS, sin CRM real y sin credenciales para probarlo), con un punto de extensión limpio para poder sustituirlo por un backend real más adelante.
Por qué un adaptador y no una integración directa
La lógica de negocio de cada conector (validación de esquema, qué se considera un error, qué campos entran y salen) se sitúa en src/connectors/*.ts y nunca accede directamente un SDK de backend concreto. En su lugar, llama a una de las cinco pequeñas interfaces de src/adapters/types.ts — CrmAdapter, TicketingAdapter, SchedulingAdapter, KnowledgeAdapter, TaskAdapter. Hoy todas están implementadas en memoria (src/adapters/mock.ts). Mañana, conectar un backend real de Salesforce/ServiceNow/Connect-Cases/Bedrock-Knowledge-Base implica escribir una nueva clase de adaptador y cambiar una única sentencia switch (src/adapters/index.ts): cero cambios en el código del conector, en sus pruebas o en las definiciones de herramientas MCP que ve un agente. Consulta docs/adding-a-real-adapter.md.
MCP client (Claude Desktop, Bedrock AgentCore, ...)
│ tool call: customer_lookup({ customerId: "cust-1001" })
▼
src/server.ts (stdio) or src/http-server.ts (deployed)
│ registers every connector as an MCP tool
▼
src/connectors/customer-lookup.ts
│ validates input (zod), calls adapters.crm.lookupCustomer(...)
▼
src/adapters/index.ts → mock.ts today, a real adapter laterRelated MCP server: CRM Agent Tools
Estructura del proyecto
src/
connectors/ one file per MCP tool — schema + business logic only
adapters/ the CRM/ticketing/scheduling/knowledge/task interfaces,
plus the in-memory mock implementations
mock-data/ seed data the mock adapters serve
mcp-server-factory.ts registers every connector on an McpServer instance
server.ts stdio entrypoint (local dev, Claude Desktop, Claude Code)
http-server.ts Streamable-HTTP entrypoint (deployed, e.g. behind Lambda)
tests/ one test file per connector + a registry sanity test
deploy/ Dockerfile + AWS SAM template for a Lambda deployment
docs/ per-connector reference + how to add a real adapterEjecutarlo localmente
npm install
npm test # 29 tests, all against the mock adapters
npm run typecheck # strict TypeScript, zero `any` in connector code
npm start # stdio MCP server — Ctrl+C to stop
npm run start:http # HTTP MCP server on :8080 (curl http://localhost:8080/health)
npm run smoke # real MCP client over stdio, calls all 7 tools end to end
npm run smoke:http # same, over HTTP — start:http must be running firstConectarlo en Claude Desktop / Claude Code
Apunta un cliente MCP al servidor stdio, p. ej., en claude_desktop_config.json de Claude Desktop:
{
"mcpServers": {
"serve7-connectors": {
"command": "npx",
"args": ["tsx", "/absolute/path/to/serve7-mcp-connectors/src/server.ts"]
}
}
}Luego pregúntale algo como «busca al cliente cust-1001» o «¿qué artículos de conocimiento tenemos sobre los avisos de cierre de emergencia?» — llamará a customer_lookup / retrieve_knowledge con los datos simulados.
Desplegarlo para un uso real (p. ej., desde Bedrock AgentCore)
Consulta deploy/README.md: crea una Lambda basada en imagen de contenedor detrás de una Function URL autenticada con IAM mediante sam deploy.
Añadir un octavo conector
Crea
src/connectors/my-new-tool.tscondefineConnector(): copia la forma de uno existente, p. ej.,create-task.ts.Añádelo al array
CONNECTORSensrc/connectors/index.ts.Añade un archivo de prueba en
tests/.Documéntalo en
docs/connectors.md.
server.ts, http-server.ts y la prueba del registro lo detectan automáticamente: no hay que tocar nada más.
Qué queda deliberadamente fuera del alcance de esta primera versión
Según el encargo original del proyecto, cada conector debía cubrir también el enfoque de autenticación (hecho, en la capa de adaptador), una plantilla de despliegue (hecho, deploy/) y documentación (hecho, docs/) — todo ello se ha incluido. Aún no está construido, y merece una continuación rápida: un adaptador real para al menos un backend (para que esto deje de ser todo simulado) y conectar un agente de Bedrock AgentCore real para que lo llame a través del endpoint HTTP desplegado, de extremo a extremo.
Available Tools
7 toolscreate_support_ticketCreate Support TicketA
File a new support ticket for a customer when their issue can't be resolved in the current interaction. Use retrieve_status afterward to check progress on a ticket already created.
Inputs: subject (required), customerId (required), description (required), priority.
create_taskCreate TaskA
Create a follow-up task for a human teammate — e.g. 'call customer back with a refund confirmation' — when something needs manual handling after the AI-driven part of the interaction ends.
Inputs: title (required), assignee (required), description (required), dueDate.
customer_lookupCustomer LookupA
Look up a customer record by customer ID, email, or phone number. Returns account status and tier so an agent (human or AI) can personalize the interaction. Call this first in nearly every flow.
Inputs: email, phone, customerId.
retrieve_knowledgeRetrieve KnowledgeA
Search the knowledge base for articles relevant to a customer's question. Returns the top matches with a relevance score and source URL so responses can be grounded and cited.
Inputs: query (required), topK.
retrieve_statusRetrieve Ticket StatusA
Check the current status of a previously created support ticket by its ticket ID. Use this before telling a customer their issue is still 'in progress' vs resolved.
Inputs: ticketId (required).
schedule_appointmentSchedule AppointmentB
Book an appointment (call-back, in-branch visit, technician visit, etc.) for an existing customer.
Inputs: startTime (required), customerId (required), serviceType (required), durationMinutes.
update_crm_recordUpdate CRM RecordA
Patch fields on an existing customer's CRM record (e.g. after verifying a new phone number or email during a call). Only include the fields you want changed.
Inputs: customerId (required), tier, email, phone, +1 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 7 tools
Each tool targets a clearly distinct resource and action: customer lookup, ticket creation/status, appointment scheduling, CRM patching, knowledge search, and task creation. There is no meaningful overlap or ambiguity between tool purposes.
Most tools follow the verb_noun pattern (create_support_ticket, retrieve_status, schedule_appointment, update_crm_record, retrieve_knowledge, create_task). The exception is customer_lookup, which reverses the order to noun_verb, creating a minor consistency deviation.
Seven tools is a well-scoped size for a customer support/CRM connector. Each tool covers a distinct workflow step, and none feel redundant or unnecessary.
The surface covers customer lookup, ticket creation and status, knowledge retrieval, and follow-up tasks, but ticket lifecycle is incomplete (no update or close) and appointment management lacks cancellation/rescheduling. These are notable gaps for a full customer support workflow.
Maintenance
Related MCP Connectors
Your team's custom tools and skills in every agent, through one MCP connection.
Unified MCP Server is a remote MCP connector for AI agents and vertical AI products that provides access to 22,000+ authorized SaaS tools across 400+ integrations and 24 categories directly inside LLMs (Claude, GPT, Gemini, Cohere). Tools operate only on explicitly authorized customer connections, enabling agents to safely read and write against live third-party systems.
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Build and run AI workflows, agents, data tables and 1000+ API integrations from any MCP client.
Related MCP Servers
- AlicenseAqualityFmaintenanceProvides 14 MCP tools for AI agent infrastructure, enabling knowledge base queries, skill search, handoffs, blueprint validation, trust scoring, identity verification, SLA validation, and compliance checks.2233 PyPIMIT
- FlicenseNot gradedqualityDmaintenanceProvides Claude agents with CRM contact lookup, action logging, and prompt cost auditing tools. Includes a Streamlit UI to demo the same tools without an MCP client.-
- AlicenseNot gradedqualityDmaintenanceEnables an AI to perform customer support workflows by looking up customers, retrieving orders, and creating support tickets through MCP tools.ISC
- FlicenseNot gradedqualityDmaintenanceProvides MCP tools for hybrid knowledge base search, grounded Q&A with citations, agent execution, and ticket/account lookups.-