BotUyo MCP Server
@botuyo/mcp — BotUyo MCP 服务器
直接从你的 AI 编码工具中创建和管理 BotUyo AI 智能体 — 无需克隆,无需构建。
快速开始
1. 获取你的 API Key
前往 admin.botuyo.com → 你的租户 → Settings → API Keys 并复制你的 pk_live_... 密钥。
2. 添加到你的 MCP 客户端
所有 兼容 MCP 的工具 配置方式相同:
{
"mcpServers": {
"botuyo": {
"command": "npx",
"args": ["-y", "@botuyo/mcp"],
"env": {
"BOTUYO_API_KEY": "pk_live_your_key_here"
}
}
}
}配置文件存放位置:
工具 | 配置文件 |
Antigravity (VS Code) | VS Code 设置 → MCP → 添加服务器 |
Claude Desktop |
|
Cursor | 项目中的 |
Windsurf |
|
Continue |
|
3. 开始构建智能体
连接后,你可以向 AI 助手询问如下内容:
“为我的租户创建一个名为 'Soporte' 的新智能体” “将 Mar 的目标更新为专注于销售” “添加一个用西班牙语问候客人的 welcomeStage” “在 Mar 上启用 EscalateToHuman 和 SearchKnowledgeBase 工具” “发布该智能体”
Related MCP server: Agent5ive MCP Server
可用工具
工具 | 描述 | 所需角色 |
| 列出所有智能体 | viewer+ |
| 获取完整智能体配置 | viewer+ |
| 渠道状态 + 管理链接 | viewer+ |
| 创建新智能体 | developer+ |
| 更新身份(语气、语言等) | developer+ |
| 创建/编辑对话阶段 | developer+ |
| 列出租户可用工具 | viewer+ |
| 在智能体上启用/禁用工具 | developer+ |
| 发布或取消发布智能体 | developer+ |
角色
角色 | 读取 | 写入 | 发布 |
| ✅ | ✅ | ✅ |
| ✅ | ✅ | ✅ |
| ✅ | ✅ | ✅ |
| ✅ | ❌ | ❌ |
渠道集成
连接渠道(WhatsApp、Instagram、Telegram、Web)必须在 admin.botuyo.com 上完成 — 它们需要交互式流程(二维码扫描、OAuth 等)。
使用 get_agent_status 查看哪些渠道已连接。它会为任何缺失的渠道返回一个指向管理面板的直接链接。
从源码构建
git clone https://github.com/MarcoAR1/botuyo-mcp.git
cd botuyo-mcp
npm install
npm run build使用 MCP Inspector 测试
BOTUYO_API_KEY=pk_live_... npm run inspect贡献
我们欢迎贡献!请参阅 CONTRIBUTING.md 获取指南。
请注意,本项目遵循 行为准则。
安全
如需报告漏洞,请参阅 SECURITY.md。
许可证
MIT © BotUyo
Available Tools
30 toolsassociate_knowledge_to_agentA
Associates (links) knowledge base document IDs to an agent. Replaces the current list of knowledgeDocumentIds on the agent. Use list_knowledge_documents to get valid document IDs.
Inputs: agentId (required), documentIds (required).
configure_agent_toolA
Crea o actualiza la config de un tool para un agente. Si la tool no está en enabledTools, la agrega automáticamente.
Dos modos:
Single-instance: configura una tool existente (ej.
send_email) con params fijos.Multi-instance: crea una instancia virtual con nombre propio basada en una tool real (requiere
baseTool). Ej:webhook_pedidosbasada encall_webhook.
Inputs: agentId (required), toolName (required), baseTool, params, +2 more.
create_agentA
Create a new agent for the tenant. The agent is created in "draft" status. Use update_agent to configure identity, stages, and tools afterward. Requires role: owner, admin, or developer.
Inputs: name (required), description, type.
create_from_templateA
Create a new agent from a pre-built template.
Templates contain pre-configured stages, connections (graph flow), tools, and channel overrides for specific industries. Business data (name, phone, email, services) is automatically hydrated into the template placeholders.
Use list_templates first to see available templates.
Requires role: owner, admin, or developer.
Inputs: templateId (required), businessName (required), contactPhone, contactEmail, +2 more.
delete_agentA
Delete an agent (soft delete). This action is IRREVERSIBLE — the agent will be deactivated and hidden.
⚠️ IMPORTANT: This tool requires explicit confirmation. You MUST set confirm to true AND provide the exact agent name in confirmName to proceed.
Requires role: owner, admin, or developer.
Inputs: agentId (required), confirm (required), confirmName (required).
delete_knowledge_documentB
Deletes a knowledge base document and all its chunks from the current tenant.
Inputs: documentId (required).
example_agentA
Returns a complete, documented example of an agent configuration JSON with ALL possible fields explained.
Use this as a reference when creating or editing agents. Every field includes a description of what it does, its type, valid values, and conditions. This is a read-only reference tool — it doesn't create or modify anything.
Useful for:
New users learning the agent config structure
Checking available fields before using import_agent_json
Understanding how stages, connections, channelFlows, and tools work together
Learning the widget theming system (cssVariables, darkCssVariables, animations, effects)
Understanding per-tool configuration (toolConfigs) with single and multi-instance patterns
Inputs: no named parameters.
export_agent_jsonA
Export the full agent configuration as clean, editable JSON.
Returns the complete agentConfig including identity, stages, connections (graph edges), channelFlows, enabledTools, channels, widgetConfig, voice, and all other settings. Also includes apiKey (read-only) for external integrations.
The exported JSON is automatically saved to a local file for easy editing:
Default: saves to ./agents/{agent-name}.json in the current working directory
Custom: use savePath to specify a different directory
Use import_agent_json to re-import after editing.
Inputs: agentId (required), savePath.
get_agentA
Get the full configuration of a specific agent, including identity (tone, language, objective, customInstructions), stages, and enabled tools.
Inputs: agentId (required).
get_agent_statusA
Get the channel connection status of an agent (WhatsApp, Instagram, Telegram, Web). Shows which channels are connected and provides a direct link to the admin panel to connect missing channels.
Inputs: agentId (required).
get_tool_configC
Ver el config completo de un tool específico de un agente (formato JSON).
Inputs: agentId (required), toolName (required).
get_tools_catalogC
Lista el catálogo completo de tools disponibles con su metadata: configurable, allowMultiInstance, configSchema, requiredIntegrations.
Inputs: filter.
import_agent_jsonA
Import/replace an agent's FULL configuration from a local JSON file or from a JSON object.
This is a FULL REPLACE operation — the entire agentConfig is overwritten with the provided config. For partial updates (changing just one field), use update_agent instead.
Null handling:
Fields set to null are stripped before saving (field is removed from the stored config)
Omitted fields are NOT preserved — this is a full replace, not a merge
Internal fields (model, temperature, summaryThreshold, voice.liveModel) are preserved automatically
Preferred workflow:
Use export_agent_json to save the agent config to a local file (auto-saved to ./agents/)
Edit the local file as needed
Use this tool with filePath pointing to that file to apply the changes
You can provide the config either via:
filePath: path to a local JSON file (preferred — reads the file from disk)
agentConfig: inline JSON object (fallback if no file)
If both are provided, filePath takes priority.
Requires role: owner, admin, or developer.
Inputs: agentId (required), filePath, agentConfig.
ingest_knowledge_urlA
Add a URL to the knowledge base. BotUyo will scrape the page content and index it for intelligent search. The document can then be associated to agents using associate_knowledge_to_agent.
Inputs: url (required), name, description, tags, +1 more.
list_agentsA
Lists all agents for the authenticated tenant. Returns id, name, description, status (draft/published), and how many tools each agent has enabled.
Inputs: no named parameters.
list_agent_versionsA
List version history for an agent. Returns all saved snapshots with version number, change source, and timestamp. Useful to see what changed and when before deciding to restore.
Inputs: agentId (required).
list_available_toolsA
List all tools available for this tenant. Returns both "core" tools (available to all tenants) and "tenant" tools (custom tools specific to your tenant). Use tool IDs when enabling tools with update_enabled_tools or configuring stage tools.
Inputs: no named parameters.
list_avatarsB
Lista los avatares 3D gratuitos disponibles en el catálogo. Muestra nombre, formato y categoría.
Inputs: no named parameters.
list_knowledge_documentsC
Lists all knowledge base documents for the current tenant. Returns document id, name, type, status, chunks count, and file size.
Inputs: limit, offset.
list_templatesA
List all available agent templates for quick agent creation.
Returns templates organized by industry vertical (beauty, clinic, restaurant, fitness, etc.). Each template includes pre-configured stages, connections (graph flow), tools, and channel overrides.
Use create_from_template to create an agent from any template.
Inputs: no named parameters.
list_tool_configsC
Lista las configs de tools de un agente agrupadas: configuradas (single y multi-instance), habilitadas sin config, y resumen de multi-instance.
Inputs: agentId (required).
publish_agentA
Publish or unpublish an agent. Published agents are live and handle incoming conversations. Unpublished agents are in "draft" status.
Requires role: owner, admin, or developer.
Inputs: agentId (required), publish (required).
remove_tool_configB
Elimina la configuración de un tool. Si es multi-instance, también la remueve de enabledTools automáticamente.
Inputs: agentId (required), toolName (required).
restore_agent_versionA
Restore an agent to a previous version (rollback). The current config is automatically saved as a new snapshot before restoring, so you can always undo a restore. Use list_agent_versions first to see available versions.
Requires role: owner, admin, or developer.
Inputs: agentId (required), version (required).
select_avatarC
Selecciona un avatar 3D para un agente. Puede ser del catálogo (por ID) o una URL custom (.glb/.vrm).
Inputs: agentId (required), avatarId, avatarUrl.
switch_tenantA
Switch the active tenant for this MCP session. If no tenantId is provided, lists all available tenants (including sub-tenants within organizations). If a tenantId is provided, switches to that tenant or sub-tenant. The switch takes effect immediately for subsequent tool calls.
Inputs: tenantId.
update_agentA
Update an agent's configuration with partial merge. Only the fields you provide are changed — everything else is preserved.
Field behavior:
Omitted field → not touched (keeps existing value)
Field with value → sets/overwrites that field
Field set to null → deletes that field from the agent config
Nested objects (identity, widgetConfig, voice, widgetConfig.cssVariables) are deep-merged: update_agent({ widgetConfig: { logoUrl: "https://..." } }) → only logoUrl changes, all other widgetConfig fields preserved
Arrays (channels, enabledTools) are replaced entirely: update_agent({ channels: ["web", "whatsapp"] }) → replaces the full channels array
Use this to configure WHO the agent is and HOW it looks. For defining WHAT the agent does in each conversation flow, use upsert_stage. For changing which tools the agent can call, use update_enabled_tools.
Requires role: owner, admin, or developer.
Inputs: agentId (required), name, description, identity, +4 more.
update_enabled_toolsA
Set which tools are enabled for an agent. This REPLACES the current list of enabled tools.
Use list_available_tools to see the valid tool IDs for your tenant first.
Requires role: owner, admin, or developer.
Inputs: agentId (required), toolIds (required).
upload_agent_mediaA
Upload a local image file to the CDN and assign it to an agent as avatar or logo.
Reads the file from the local filesystem, uploads it via the backend media API (Cloudinary), and updates the agent's widgetConfig with the resulting URL.
Supported formats: webp, png, jpg, jpeg, svg. Max size: 2MB.
Inputs: agentId (required), filePath (required), mediaType.
upsert_stageA
Create or update a named stage in an agent's conversation flow.
A stage defines what the agent does at a particular step in the conversation. Stages are nodes in a graph connected by edges (connections).
Examples of stage names: "welcomeStage", "salesStage", "hostStage", "checkoutStage"
The stage config is merged with existing stages — existing stages you don't specify are kept unchanged. Connections and channelFlows replace existing values when provided.
Requires role: owner, admin, or developer.
Inputs: agentId (required), stageName (required), stageConfig (required), connections, +1 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 30 tools
Most tools have distinct purposes, but some overlap exists that could cause confusion. For example, 'configure_agent_tool' and 'update_enabled_tools' both relate to tool management, and 'list_available_tools' vs 'get_tools_catalog' serve similar listing functions. Descriptions help clarify, but an agent might misselect between these pairs.
Naming conventions are mixed, with some tools using verb_noun patterns (e.g., 'create_agent', 'list_agents') and others using noun_verb or more descriptive phrases (e.g., 'associate_knowledge_to_agent', 'example_agent'). There's also inconsistency in language, with some tools in Spanish and others in English, which adds to the lack of a predictable pattern.
With 30 tools, the count feels excessive for an agent management server, leading to potential overwhelm and redundancy. While the domain is broad, many tools could be consolidated (e.g., multiple listing tools), making the set feel heavy and less focused than ideal for efficient agent use.
The tool set provides comprehensive coverage for agent lifecycle management, including creation, configuration, publishing, and version control. It also handles knowledge bases, templates, and media. Minor gaps might exist in advanced analytics or bulk operations, but core workflows are well-supported with no dead ends.
Maintenance
Related MCP Connectors
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Free public MCP for AI agents — 193 tools, 44 workflows. No API key.
Your org's AI agents, tasks, runs, search, and brain files as MCP tools and resources.
Build and manage AI-native customer support agents from Claude or any MCP client.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables management of Mistral AI agents through MCP tools including creating, listing, searching, viewing details, and deleting agents. Integrates with Mistral API to provide agent management capabilities through natural language interactions.1-
- AlicenseNot gradedqualityDmaintenanceEnables MCP-compatible clients to interact with deployed Agent5ive agents as tools. Allows users to query agent purposes and send messages to leverage Agent5ive capabilities through natural language.MIT
- AlicenseAqualityDmaintenanceExposes Azure AI Foundry agents, workflows, and AI Search vector-database capabilities as MCP tools, enabling natural language interaction with agents, semantic search, and index management.102MIT
- FlicenseNot gradedqualityDmaintenanceA universal MCP hub that connects any AI agent to any tool, enabling agent interoperability with 50+ pre-built tools, custom tool creation, and tool discovery.-