markovo-mcp
Enables converting supported files and explicitly authorized public HTTPS pages into clean, structured Markdown through the Markovo service, with file access restricted to a configured root directory.
README
Markovo CLI and MCP server
Markovo turns supported files and explicitly authorized public HTTPS pages into clean, structured Markdown. For the dedicated web experience and product details, visit the PDF to Markdown Converter.
This public repository is the distribution source for the customer-side
markovo CLI and the markovo-mcp stdio MCP server. It contains no local
conversion engine, backend service, deployment configuration, production
credentials, or billing implementation. Conversion runs through the
account-metered service at https://markovo.net.
Install
pip install "markovo==0.1.1"
Or run it without a permanent installation:
uvx --from "markovo==0.1.1" markovo --help
uvx --from "markovo==0.1.1" markovo-mcp
API key and Credits
- Create an API key in the Developer portal.
- Store it in your shell or MCP host secret store; do not paste it into prompts or commit it to source control.
- Verify the setup:
export MARKOVO_API_KEY="your-key"
markovo doctor --json
Every conversion requires an explicit Credit ceiling. When a key is missing, the client points to the Developer portal. When Credits are insufficient, the service response points to the billing page so the account can be topped up before retrying.
MCP configuration
Remote MCP (recommended for hosted agents)
Connect the Streamable HTTP endpoint:
https://markovo.net/mcp
The MCP host discovers Markovo's OAuth 2.1 metadata, opens the Markovo
account sign-in and consent screen, and stores its own scoped token. You
do not create, paste, or share an API key for this remote connection.
Remote tools can inspect capabilities, Credits, and owned jobs; convert
one explicitly supplied public HTTPS page with a required
max_credit_units ceiling; and create or revoke short-lived result-image
links. Remote MCP cannot read local files. Use the stdio package below or
REST multipart upload when a local file must be converted.
Markovo never tops up Credits or changes a plan automatically. A low balance response points to https://markovo.net/app#billing and waits for the account owner to act.
Local stdio MCP
{
"mcpServers": {
"markovo": {
"command": "uvx",
"args": ["--from", "markovo==0.1.1", "markovo-mcp"],
"env": {
"MARKOVO_API_KEY": "${MARKOVO_API_KEY}",
"MARKOVO_MCP_ROOT": "/path/to/safe/project"
}
}
}
}
MCP clients should inject MARKOVO_API_KEY through their secret or
environment-variable settings. The key and service origin are never MCP
tool arguments, so prompts cannot redirect the credential. The public
client accepts only https://markovo.net and does not follow authenticated
redirects.
The MCP server can read and write only within the required
MARKOVO_MCP_ROOT. Set it to the smallest dedicated project directory
your MCP host needs. If it is missing, the server refuses all file tools;
parent-directory and symlink escapes are rejected.
export MARKOVO_MCP_ROOT="/path/to/safe/project"
Structured errors point to the Developer portal when a key is missing and to Billing when Credits are insufficient.
Links
- Markovo home
- PDF to Markdown
- Documentation
- Remote MCP guide
- Pricing and Credits
- Developer portal and API keys
- Billing and top-up
Security
Report security issues privately using the instructions in SECURITY.md. Do not include API keys, customer documents, or private conversion results in a public issue.
<!-- mcp-name: io.github.fisher-byte/markovo -->
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。