使用说明

Arvore MCP Servers

A collection of Model Context Protocol (MCP) servers developed by Arvore for seamless integration with AI assistants like Claude.

📦 Packages

This monorepo contains the following MCP servers:

@arvoretech/aws-secrets-manager-mcp

Manage AWS Secrets Manager secrets directly from your AI assistant.

Features:

Create, read, update, and delete secrets
List all secrets in your account
Describe secret metadata

@arvoretech/datadog-mcp

Query and analyze Datadog monitoring data.

Features:

Query metrics and time series data
Search logs
Get service maps and APM data
List hosts and active metrics
Search traces

@arvoretech/mysql-mcp

Execute read-only MySQL queries safely.

Features:

Execute SELECT queries
List tables in database
Read-only operations for safety

@arvoretech/postgresql-mcp

Execute read-only PostgreSQL queries safely.

Features:

Execute SELECT queries
List tables, schemas, and databases
Describe table structure
Full schema support
Read-only operations for safety

@arvoretech/clickhouse-mcp

Execute read-only ClickHouse queries safely.

Features:

Execute SELECT queries
List tables and databases
Describe table structure (columns, types, keys)
Read-only operations for safety

@arvoretech/npm-registry-mcp

Search and get information about npm packages.

Features:

Get package information and metadata
Get download statistics
Search packages by query

@arvoretech/launchdarkly-mcp

Manage LaunchDarkly feature flags directly from your AI assistant.

Features:

List, search, create, and delete feature flags
Toggle flags ON/OFF per environment
Add targeting rules (in, contains, segmentMatch, etc.)
List and inspect segments
Multi-environment support (production, test, etc.)

@arvoretech/tempmail-mcp

Self-hosted temporary email server with SMTP + SQLite.

Features:

Built-in SMTP server to receive real emails
Create and manage temporary email accounts
Read, list, and delete emails
SQLite storage — no third-party dependencies
Configurable domain via AWS Route 53

@arvoretech/google-chat-mcp

Google Chat integration for managing spaces, members, and messages.

Features:

List and inspect Google Chat spaces
List space members
Read message history with filtering and pagination
Send messages (with thread support)
Delete messages
Service account JWT auth (no external dependencies)
Domain-wide delegation for full user-level access

@arvoretech/gmail-mcp

Gmail integration — read, send, and manage messages with OAuth user flow.

Features:

Search messages with full Gmail query syntax (from:, is:unread, newer_than:, etc.)
Read messages and full threads with parsed headers and body
Create drafts (always available, even with sending disabled)
Send mail and reply with proper threading (In-Reply-To, References, threadId)
Modify labels (mark read, archive, star, trash)
OAuth Desktop client flow with browser-based login
Encrypted token storage (AES-256-GCM + macOS keychain)
Send-mode opt-in via GMAIL_MCP_ALLOW_SEND to prevent accidental sends

@arvoretech/google-slides-mcp

Google Slides integration — create, edit and visually inspect presentations.

Features:

Search decks in Drive, create from scratch or copy a template
Compact deck summary with object ids, placeholders and speaker notes
Screenshot any slide as PNG so the model can verify layout visually
Typed edits: add slide from layout, insert/replace text, insert image, speaker notes
Raw batchUpdate escape hatch for full Slides API styling and geometry
Export as pdf / pptx / txt
Runs over stdio or streamable-http

@arvoretech/google-docs-mcp

Google Docs integration — create, read and edit documents.

Features:

Search documents in Drive, create from scratch or copy a template
Read a document as Markdown; map it as indexed blocks when editing
Whole formatted document from Markdown in one call, no index math
Typed edits: append styled paragraphs, insert text, tables, images, page breaks
Text and paragraph formatting: styles, colors, links, alignment, bullets
Raw batchUpdate escape hatch for the full Docs API
Export as pdf / docx / txt / md / html / odt / rtf / epub
Runs over stdio or streamable-http

@arvoretech/google-tasks-mcp

Google Tasks integration — the to-do list of Gmail and Google Calendar, which has no official MCP.

Features:

List, create, rename and delete task lists
List open or completed tasks, filtered by due date, completion date or last change
Create tasks with notes, due date and subtasks; update, complete, reopen and delete them
Reorder, nest and move tasks between lists
Per-person Google sign-in for claude.ai connectors
Runs over stdio or streamable-http

@arvoretech/google-sheets-mcp

Google Sheets integration — read, write, reshape and format spreadsheets.

Features:

Search spreadsheets in Drive, create from scratch or copy a template
Describe every tab with its gid, size, named ranges and a preview of the first rows
Read and write A1 ranges, with formulas and locale-aware parsing
Append rows to a table without computing where the data ends
Manage tabs, insert and delete rows and columns, sort, find and replace
Cell formatting: fonts, colors, alignment, wrapping, number formats
Raw batchUpdate escape hatch for charts, borders, validation and filters
Export as pdf / xlsx / csv
Runs over stdio or streamable-http

@arvoretech/memory-mcp

Team memory — persistent knowledge base with semantic search for AI-assisted development.

Features:

Semantic search across team memories using local embeddings
Multilingual support (Portuguese + English) via paraphrase-multilingual-MiniLM-L12-v2
Categories: decisions, conventions, incidents, domain, gotchas
Create, archive, and remove memories via MCP tools
Embedding cache for fast startup
Keyword fallback when embeddings are unavailable

Runtime Lens — Runtime inspection with inline values for React, NestJS and Next.js. Two components in one package: a VS Code/Cursor/Kiro extension with inline value display and an MCP server for AI-assisted debugging.

Features:

Inline runtime values displayed in the editor
Runtime agent injected via --require that patches console.log/warn/error/debug
WebSocket-based communication between agent and extension
14 MCP tools for AI-assisted debugging: log tailing, regex search, error inspection, HTTP request inspection, performance metrics, and more
Framework detection for React, NestJS, and Next.js
Real-time console capture with start/stop control
Permanent setup via ~/.zshrc — no need to inject every session

@arvoretech/meet-transcriptions-mcp

Semantic search across meeting transcriptions generated by Meet Transcriber.

Features:

Semantic search using local multilingual embeddings (paraphrase-multilingual-MiniLM-L12-v2)
LanceDB vector store with keyword fallback
Filter by speaker
List and retrieve full transcription content
Reads .md files from ~/Documents/meet-transcriptions/ by default

@arvoretech/sendgrid-mcp

Manage SendGrid Dynamic Templates directly from your AI assistant.

Features:

List, create, update, and delete dynamic templates
Create and manage template versions (HTML content, subject, plain text)
Activate specific template versions
Full CRUD for templates and versions via SendGrid API v3

@arvoretech/await-until-mcp

Polling/waiting primitives — repeatedly check conditions until they're met or time out.

Features:

Execute shell commands until output matches a condition
Poll HTTP endpoints until they return expected status/body
Watch filesystem until files exist, disappear, or contain content
Call any other MCP server's tool and poll until the result matches (auto-discovers from mcp.json)
Configurable interval, timeout, and match conditions

@arvoretech/mcp-proxy

MCP Proxy Gateway — consolidates N upstream MCP servers behind 2 tools.

Features:

Exposes only mcp_search and mcp_call to the agent
Semantic search using multilingual embeddings (paraphrase-multilingual-MiniLM-L12-v2)
Hybrid ranking: BM25 lexical + cosine similarity
Output shaping: strips redundant fields, truncates text, assigns short refs
Cursor-based pagination with TTL
Supports stdio and HTTP transports with env var expansion

@arvoretech/agent-teams-lead-mcp

Agent Teams Lead — spawn and coordinate multiple AI agent sessions working together as a team. Inspired by Anthropic's agent teams for Claude Code, built as an editor-agnostic MCP layer.

Features:

Spawn teammates as independent CLI processes (Kiro, Cursor, Claude Code, OpenCode)
Shared task list with dependencies and exclusive file paths
Inter-agent messaging (direct, broadcast, or to lead)
Atomic mkdir-based file locking for safe parallel task claiming
Audit log in .agent-teams/team.log

@arvoretech/agent-teams-teammate-mcp

Agent Teams Teammate — MCP injected into each teammate session for claiming tasks, messaging, and publishing artifacts.

Features:

Claim pending tasks with dependency and lock validation
Communicate with other teammates and the lead
Publish artifacts (markdown, JSON, or code) linked to tasks
Automatically injected by the lead — no manual configuration needed

@arvoretech/agent-teams-chat-mcp

Agent Teams Chat — cross-developer agent communication via Slack threads. Enables AI agents from different developers to communicate asynchronously through a shared Slack channel.

Features:

Open conversation threads in a team Slack channel
Reply to existing threads
Read thread messages and catch up on conversations
List recent threads with reply counts and participants
Search threads by topic or content
Configurable message format using handlebars-style templates

@arvoretech/agent-teams-ui

Agent Teams UI — TUI dashboard for monitoring agent teams in real time. Watches .agent-teams/ and auto-refreshes.

Features:

Team view with per-teammate task counts and current activity
Kanban board with task selection and detail view
WhatsApp-style message view with date separators and sender grouping
Chat view showing agent stdout as conversation, filterable by agent
Live timer, progress bar, and new-activity indicators per tab
Auto-refresh via filesystem watching (chokidar)

@arvoretech/kanban-mcp

Persistent kanban boards for AI agents — manage tasks across sessions with multi-agent coordination, semantic search, and a Linear-style web UI.

Features:

12 MCP tools for board and card management
Multi-session: claim/release cards with session tracking across parallel chats
Semantic search via LanceDB with multilingual embeddings
Subtasks via parent_card_id
Optional web UI (Linear-style dark theme with drag-and-drop, hotkeys, context menus)
JSON persistence with atomic writes
Re-read from disk for multi-process visibility

@arvoretech/metabase-mcp

Interact with Metabase BI platform directly from your AI assistant.

Features:

CRUD for cards (questions), dashboards, and collections
List databases and tables
Execute native SQL queries
Run saved card queries
Add cards to dashboards
Get collection items

@arvoretech/mgc-mcp

Manage Magalu Cloud infrastructure through the MGC CLI.

Features:

Virtual Machines: list, create, start, stop, reboot, delete
Object Storage: buckets and objects management
Networking: VPCs, subnets, public IPs
Kubernetes: cluster management and kubeconfig retrieval
Block Storage and DBaaS instance listing
Semantic search across Magalu developer documentation
Configurable region via MGC_REGION env var

@arvoretech/arc-browser-mcp

Control Arc Browser on macOS via a hybrid AppleScript + Chrome DevTools Protocol approach.

Features:

Tab management: list, open, close, search, focus
Space management: list, get active, switch
Screenshots via CDP (viewport or full page)
Network request capture and console log monitoring
DOM interaction: click, hover, type, scroll, wait for selector
JavaScript execution (AppleScript for quick queries, CDP for async/await)
Cookie inspection and page content extraction

@arvoretech/posthog-mcp

PostHog analytics, feature flags, experiments, and more. Works with both PostHog Cloud and self-hosted instances.

Features:

HogQL queries and structured queries (Trends, Funnels, Retention, Lifecycle, Paths, Stickiness)
Feature flags: create, update, list, delete with rollout control
A/B test experiments with variants and metrics
Dashboards and saved insights management
Event and property definitions exploration
Persons search and cohort management
Surveys, early access features, and annotations
Cross-entity search

@arvoretech/slack-advanced-mcp

Advanced Slack integration using user tokens -- messages sent as the authenticated user.

Features:

Fuzzy user search with disk cache (47k+ users load instantly)
Smart DM: resolve user by name/email/ID, auto-open DM channel
Writing style analysis (emoji usage, punctuation, vocabulary, formality)
Thread extraction from Slack URLs with participant resolution
Audio transcription via ElevenLabs STT
Image download and base64 return for model analysis
Rate limit retry with backoff

@arvoretech/whatsapp-mcp

Drive a single WhatsApp account from your AI agent using Baileys.

Features:

Pair via QR code (ASCII in terminal, PNG file, or base64 data URL)
Send and receive text, images, audio (PTT or normal), video, and documents
Local SQLite store for chats, messages, and contacts so the agent can recall past conversations
Brazilian phone number normalization with WhatsApp-side validation
Set typing presence, mark messages as read, react, edit, and delete
Single connection — pair one WhatsApp account from your local agent

@arvoretech/leaf-mcp

Read-only access to Leaf, Arvore's collaborative document editor.

Features:

Full-text search over document titles and bodies
Read any document as markdown (headings, lists, tables, code, images)
Browse Notion-style databases: typed properties, views, and rows with resolved option names
List organizations with members and teamspaces
Read comment threads with replies, block anchors, and resolved state

🚀 Quick Start

Installation

Install individual packages:

npm install -g @arvoretech/aws-secrets-manager-mcp
npm install -g @arvoretech/datadog-mcp
npm install -g @arvoretech/mysql-mcp
npm install -g @arvoretech/postgresql-mcp
npm install -g @arvoretech/clickhouse-mcp
npm install -g @arvoretech/npm-registry-mcp
npm install -g @arvoretech/launchdarkly-mcp
npm install -g @arvoretech/tempmail-mcp
npm install -g @arvoretech/google-chat-mcp
npm install -g @arvoretech/gmail-mcp
npm install -g @arvoretech/memory-mcp
npm install -g @arvoretech/runtime-lens-mcp
npm install -g @arvoretech/meet-transcriptions-mcp
npm install -g @arvoretech/sendgrid-mcp
npm install -g @arvoretech/await-until-mcp
npm install -g @arvoretech/mcp-proxy
npm install -g @arvoretech/agent-teams-lead-mcp
npm install -g @arvoretech/agent-teams-teammate-mcp
npm install -g @arvoretech/agent-teams-chat-mcp
npm install -g @arvoretech/metabase-mcp
npm install -g @arvoretech/mgc-mcp
npm install -g @arvoretech/kanban-mcp
npm install -g @arvoretech/posthog-mcp
npm install -g @arvoretech/slack-advanced-mcp
npm install -g @arvoretech/whatsapp-mcp

Or using pnpm:

pnpm add -g @arvoretech/aws-secrets-manager-mcp

Configuration

Add to your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "aws-secrets-manager": {
      "command": "npx",
      "args": ["-y", "@arvoretech/aws-secrets-manager-mcp"]
    },
    "datadog": {
      "command": "npx",
      "args": ["-y", "@arvoretech/datadog-mcp"],
      "env": {
        "DATADOG_API_KEY": "your-api-key",
        "DATADOG_APP_KEY": "your-app-key",
        "DATADOG_SITE": "datadoghq.com"
      }
    },
    "mysql": {
      "command": "npx",
      "args": ["-y", "@arvoretech/mysql-mcp"],
      "env": {
        "MYSQL_HOST": "localhost",
        "MYSQL_USER": "readonly_user",
        "MYSQL_PASSWORD": "password",
        "MYSQL_DATABASE": "your_database",
        "MYSQL_PORT": "3306"
      }
    },
    "postgresql": {
      "command": "npx",
      "args": ["-y", "@arvoretech/postgresql-mcp"],
      "env": {
        "POSTGRESQL_HOST": "localhost",
        "POSTGRESQL_PORT": "5432",
        "POSTGRESQL_USER": "postgres",
        "POSTGRESQL_PASSWORD": "password",
        "POSTGRESQL_DATABASE": "your_database",
        "POSTGRESQL_SSL": "false"
      }
    },
    "clickhouse": {
      "command": "npx",
      "args": ["-y", "@arvoretech/clickhouse-mcp"],
      "env": {
        "CLICKHOUSE_URL": "https://your-host:8443",
        "CLICKHOUSE_USER": "default",
        "CLICKHOUSE_PASSWORD": "secret",
        "CLICKHOUSE_DATABASE": "default"
      }
    },
    "npm-registry": {
      "command": "npx",
      "args": ["-y", "@arvoretech/npm-registry-mcp"]
    },
    "launchdarkly": {
      "command": "npx",
      "args": ["-y", "@arvoretech/launchdarkly-mcp"],
      "env": {
        "LAUNCHDARKLY_EMAIL": "your-email@example.com",
        "LAUNCHDARKLY_PASSWORD": "your-password",
        "LAUNCHDARKLY_PROJECT": "default",
        "LAUNCHDARKLY_ENVIRONMENT": "production"
      }
    },
    "tempmail": {
      "command": "npx",
      "args": ["-y", "@arvoretech/tempmail-mcp"],
      "env": {
        "CLOUDFLARE_ACCOUNT_ID": "your-account-id",
        "CLOUDFLARE_D1_DATABASE_ID": "your-database-id",
        "CLOUDFLARE_API_TOKEN": "your-api-token",
        "TEMPMAIL_DOMAIN": "yourdomain.com"
      }
    },
    "google-chat": {
      "command": "npx",
      "args": ["-y", "@arvoretech/google-chat-mcp"],
      "env": {
        "GOOGLE_CHAT_CREDENTIALS_PATH": "/path/to/service-account.json",
        "GOOGLE_CHAT_USER_EMAIL": "user@yourdomain.com"
      }
    },
    "gmail": {
      "command": "npx",
      "args": ["-y", "@arvoretech/gmail-mcp"],
      "env": {
        "GMAIL_MCP_CLIENT_ID": "your-oauth-client-id",
        "GMAIL_MCP_CLIENT_SECRET": "your-oauth-client-secret",
        "GMAIL_MCP_ALLOW_SEND": "false"
      }
    },
    "team-memory": {
      "command": "npx",
      "args": ["-y", "@arvoretech/memory-mcp"],
      "env": {
        "MEMORY_PATH": "./memories"
      }
    },
    "runtime-lens": {
      "command": "npx",
      "args": ["-y", "@arvoretech/runtime-lens-mcp"],
      "env": {
        "RUNTIME_LENS_PROJECT_ROOT": "."
      }
    },
    "meet-transcriptions": {
      "command": "npx",
      "args": ["-y", "@arvoretech/meet-transcriptions-mcp"]
    },
    "sendgrid": {
      "command": "npx",
      "args": ["-y", "@arvoretech/sendgrid-mcp"],
      "env": {
        "SENDGRID_API_KEY": "SG.your-api-key"
      }
    },
    "await-until": {
      "command": "npx",
      "args": ["-y", "@arvoretech/await-until-mcp"]
    },
    "mcp-proxy": {
      "command": "node",
      "args": ["packages/mcp-proxy/dist/index.js"],
      "env": {
        "MYSQL_HOST": "localhost",
        "MCP_PROXY_UPSTREAMS": "[{\"name\":\"mysql\",\"command\":\"npx\",\"args\":[\"-y\",\"@arvoretech/mysql-mcp\"],\"env\":{\"MYSQL_HOST\":\"${MYSQL_HOST}\"}}]"
      }
    },
    "agent-teams-lead": {
      "command": "npx",
      "args": ["-y", "@arvoretech/agent-teams-lead-mcp"],
      "env": {
        "WORKSPACE_PATH": "/path/to/your/workspace"
      }
    },
    "agent-teams-chat": {
      "command": "npx",
      "args": ["-y", "@arvoretech/agent-teams-chat-mcp"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-your-token",
        "SLACK_CHANNEL": "C0123456789",
        "AGENT_IDENTITY": "Your Name"
      }
    },
    "metabase": {
      "command": "npx",
      "args": ["-y", "@arvoretech/metabase-mcp"],
      "env": {
        "METABASE_URL": "https://your-metabase-instance.com",
        "METABASE_API_KEY": "your-api-key"
      }
    },
    "mgc": {
      "command": "npx",
      "args": ["-y", "@arvoretech/mgc-mcp"],
      "env": {
        "MGC_REGION": "br-ne1",
        "MAGALU_DOCS_DIR": "./magalu-cloud-docs"
      }
    },
    "kanban": {
      "command": "npx",
      "args": ["-y", "@arvoretech/kanban-mcp"],
      "env": {
        "KANBAN_PATH": "./kanban",
        "KANBAN_UI": "true",
        "KANBAN_PORT": "4799"
      }
    },
    "posthog": {
      "command": "npx",
      "args": ["-y", "@arvoretech/posthog-mcp"],
      "env": {
        "POSTHOG_BASE_URL": "https://posthog.example.com",
        "POSTHOG_API_KEY": "phx_your_personal_api_key"
      }
    },
    "slack-advanced": {
      "command": "npx",
      "args": ["-y", "@arvoretech/slack-advanced-mcp"],
      "env": {
        "SLACK_USER_TOKEN": "xoxp-your-user-token",
        "ELEVENLABS_API_KEY": "sk_your-elevenlabs-key"
      }
    },
    "whatsapp": {
      "command": "npx",
      "args": ["-y", "@arvoretech/whatsapp-mcp"],
      "env": {
        "WHATSAPP_LOG_LEVEL": "warn"
      }
    }
  }
}

Note: These packages are published to the public npm registry and can be used directly with npx without any additional configuration.

🛠️ Development

Prerequisites

Node.js >= 20
pnpm >= 9

Setup

git clone https://github.com/arvoreeducacao/arvore-mcp-servers.git
cd mcp-servers
pnpm install

Build all packages

pnpm build

Run tests

pnpm test

Run tests with coverage

pnpm test:cov

Lint

pnpm lint

Development mode

Run a specific package in development mode:

cd packages/aws-secrets-manager
pnpm dev

📝 Adding a New Package

Create a new directory in packages/
Follow the structure of existing packages
Update the root README.md
Add package-specific documentation

🚢 Publishing

Automatic Publishing

This monorepo uses GitHub Actions for automatic publishing to npm. Publishing happens in these scenarios:

1. Push to main branch

Only the packages changed in the push are published when code is merged to main (after CI passes). The publish job detects which packages/<name> directories were touched via git diff against the previous commit and runs npm publish only for those. Remember to bump the version in the changed package's package.json before merging, otherwise npm will skip it as already published.

2. Manual workflow dispatch

You can manually trigger publishing for a specific package:

Go to GitHub Actions
Select "Publish Packages" workflow
Click "Run workflow"
Specify package name (e.g., @arvoretech/aws-secrets-manager-mcp)
Choose version bump type (major, minor, or patch)

Setup Requirements

To enable automatic publishing, add these secrets to your GitHub repository:

NPM_TOKEN: Required for publishing to npm
- Create a token at https://www.npmjs.com/settings/tokens
- Select "Automation" token type
- Add as repository secret in GitHub Settings > Secrets and variables > Actions
CODECOV_TOKEN (optional): For coverage reports
- Get from https://codecov.io
- Add as repository secret

Manual Publishing

If needed, you can publish manually (requires authentication):

# Publish all packages
pnpm -r publish

# Publish specific package
cd packages/aws-secrets-manager
npm publish

Version Management

To bump versions:

# In a specific package directory
cd packages/aws-secrets-manager
pnpm version patch  # or minor, major

🤝 Contributing

Contributions are welcome! Please read our contributing guidelines before submitting PRs.

📄 License

MIT License - see the LICENSE file for details.

🔗 Links

📞 Support

For issues and questions, please open an issue on GitHub.

Made with ❤️ by Arvore Education

使用说明

Arvore MCP Servers

📦 Packages

🚀 Quick Start

Installation

Configuration

🛠️ Development

Prerequisites

Setup

Build all packages

Run tests

Run tests with coverage

Lint

Development mode

📝 Adding a New Package

🚢 Publishing

Automatic Publishing

1. Push to main branch

2. Manual workflow dispatch

Setup Requirements

Manual Publishing

Version Management

🤝 Contributing

📄 License

🔗 Links

📞 Support

相关 MCP

AKShare One MCP 股市数据服务

实时网络搜索服务

arXiv 论文检索 MCP 服务器

移动自动化MCP服务器

AI持久记忆服务器

Git仓库管理MCP服务器