Genesys MCP
Enables AI agents to interact with Genesys Cloud platform through MCP tools, resources, and prompts, supporting queues, conversations, users, presence, and analytics with production-ready features like Streamable HTTP and per-request authentication.
README
Genesys MCP
Vendor-neutral Model Context Protocol server for Genesys Cloud.
genesys-mcp is a Model Context Protocol (MCP) server that exposes Genesys Cloud
platform capabilities — queues, conversations, users, presence, analytics — as
MCP tools, resources, and prompts. Any MCP-compatible AI agent can speak to
Genesys through it: Claude Desktop, Claude Code, Cursor, Continue, Cline, and
any custom agent built on LangGraph, LangChain, CrewAI, AutoGen, or a raw
provider SDK.
Status
v0.1.0 — pre-release. The framework, transports, and reliability layer are
in place. The first batch of read-only tools lands in v0.2; curated writes
(behind --enable-writes) and Resources/Prompts land in v0.3. The v1.0 launch
ships everything together with PyPI, Docker (GHCR), an npx shim, and an MCPB
bundle.
Why
The MCP ecosystem already has a Genesys Cloud server, but it ships stdio-only
without HTTP transport, has no per-request auth context (a single module-global
boolean gates the whole process), and has no production reliability layer
(no 429 handling, no token refresh, no structured logging). genesys-mcp is
designed for the production surface: hosted-ready Streamable HTTP, per-request
auth, 429-aware retry, refresh-on-401, structured logs, and OpenTelemetry
tracing — all from day one, all vendor-neutral.
Quickstart
Tools are not yet shipped — these snippets show the install pattern. v0.2 will list the actual tool surface here.
Claude Desktop
Add an entry under mcpServers in your claude_desktop_config.json:
{
"mcpServers": {
"genesys": {
"command": "uvx",
"args": ["genesys-mcp"],
"env": {
"GENESYS_REGION": "mypurecloud.com",
"GENESYS_CLIENT_ID": "...",
"GENESYS_CLIENT_SECRET": "..."
}
}
}
}
Claude Code or Cursor
Add an entry to your editor's MCP server config (.claude/settings.json for
Claude Code, the Cursor MCP settings panel for Cursor):
{
"mcpServers": {
"genesys": {
"command": "genesys-mcp",
"args": ["--transport", "stdio"],
"env": {
"GENESYS_REGION": "mypurecloud.com",
"GENESYS_CLIENT_ID": "...",
"GENESYS_CLIENT_SECRET": "..."
}
}
}
}
Programmatic Python
Connect via the official mcp Python client over stdio or HTTP:
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client
params = StdioServerParameters(command="genesys-mcp", args=["--transport", "stdio"])
async with stdio_client(params) as (read, write):
async with ClientSession(read, write) as session:
await session.initialize()
# tools = await session.list_tools() # available from v0.2
For HTTP transport, run genesys-mcp --transport http --port 8000 and connect
your MCP client to http://localhost:8000.
Install
pip install genesys-mcp
Coming v1.0: published on PyPI, GHCR, npm shim, and as an MCPB bundle attached to GitHub Releases.
Transports
| Transport | Use case | CLI |
|---|---|---|
stdio (default) |
Local clients — Claude Desktop, Claude Code, Cursor, npx invocation |
genesys-mcp |
http |
Networked / hosted deployment, multi-agent gateways | genesys-mcp --transport http --host 0.0.0.0 --port 8000 |
Both transports share the same tool surface. Transport selection is
environment-driven (MCP_TRANSPORT=stdio|http) or CLI-driven (--transport).
Security
The HTTP transport in v0.x ships unauthenticated. The MCP framing layer does not perform any caller authentication; the server trusts whatever process can reach the listening port. Operators have two responsible deployment options:
- Bind to
127.0.0.1only (the defaultMCP_HOST) and let local agents connect over loopback. - Bind to a non-loopback address only behind a reverse proxy (nginx, Caddy,
Cloudflare Access, an API gateway) that performs authentication and
authorisation before forwarding to
genesys-mcp.
Binding to 0.0.0.0 without a fronting auth layer exposes the upstream
Genesys credentials to anyone on the network. The Docker image therefore
defaults MCP_HOST=127.0.0.1; operators who genuinely need a public bind
must opt in explicitly with -e MCP_HOST=0.0.0.0.
A hosted, multi-tenant gateway with first-class authentication, audit, and
RBAC is on the Pro/SaaS roadmap (see ADR-006 in docs/warm/decisions.md).
Trust model
A few things to know before installing plugins or running this in production:
- Entry-point plugins (the
genesys_mcp.pluginsgroup) execute in the server process with full access to every dependency, environment variable, and Genesys credential the server has. Only install plugins you trust. - The license hook (
genesys_mcp.licensegroup) is a feature gate, not a security control. Any installed package can register one and any plugin can ignore the answer. It exists so the Pro package can express entitlements; treat it accordingly. - OAuth client secrets are loaded from environment variables (and the
.envfile if present) and the server itself never logs them. We cannot guarantee the same of downstream plugins — vet them, or run withLOG_LEVEL=INFOand inspect output before trusting them with secrets.
Configuration
Configuration is read from environment variables (or a .env file). See
.env.example for the full list. The most important variables:
| Variable | Default | Description |
|---|---|---|
MCP_TRANSPORT |
stdio |
Transport selection (stdio or http) |
MCP_HOST |
127.0.0.1 |
HTTP bind host |
MCP_PORT |
8000 |
HTTP bind port |
LOG_LEVEL |
INFO |
DEBUG / INFO / WARNING / ERROR / CRITICAL |
GENESYS_MCP_ENABLE_WRITES |
false |
Opt-in to low-blast-radius write tools |
GENESYS_REGION |
mypurecloud.com |
Genesys Cloud region (e.g. mypurecloud.de, apne2.pure.cloud) |
GENESYS_CLIENT_ID |
— | OAuth Client Credentials client id |
GENESYS_CLIENT_SECRET |
— | OAuth Client Credentials client secret |
Development
git clone https://github.com/digitalvanguardsolutions/genesys-mcp.git
cd genesys-mcp
uv sync --all-extras
uv run pytest
uv run ruff check src tests
uv run mypy src
uv run genesys-mcp --transport stdio
pre-commit install will wire up ruff, mypy, and a pre-push pytest run.
License
MIT — see LICENSE. Copyright 2026 Digital Vanguard Solutions.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。