The official MarkItDown MCP server exposes document-to-Markdown conversion through Streamable HTTP and SSE.
Launch in your Sealos workspace.
Start with the resources defined by the MarkItDown MCP template. Review CPU, memory, persistent storage, and network allocations for every service in Canvas. Capacity needs depend on your data and workload; monitor usage as they grow.
Template configuration and setupUse the application URL or connection details shown in Canvas. Follow the deployment guide for first-time account setup or client configuration, and keep generated credentials available for that step.
Sealos monthly plans include compute, memory, storage, and traffic. Size your plan for all deployed services. Software licenses and external AI or API services may have separate terms and charges. Confirm the applicable plan and optional charges in Cost Center.
Compare Sealos resource plansFor current cloud charges, refer to Sealos plan pricing. Upstream documentation may reference earlier billing models.
MarkItDown MCP is Microsoft's official MCP server for converting documents and remote resources to Markdown. This template runs the upstream Streamable HTTP and SSE server on Sealos Cloud and exposes the protocol endpoints through HTTPS.
The markitdown-mcp package exposes one MCP tool, convert_to_markdown(uri), for http:, https:, file:, and data: URIs. The official container starts the server on port 3001 with Streamable HTTP and SSE transports.
The Sealos template keeps the upstream protocol surface intact. It provisions a 1 GiB /workdir volume for files that the MCP server must read, an HTTPS Ingress for /mcp/, /sse, and /messages/, and a Canvas App entry pointing to /mcp/. The server has no built-in authentication, so access must remain limited to trusted MCP clients and network controls.
/workdir and reference them with file: URIs./mcp/.The template includes the Docker MCP Catalog image built from Microsoft's upstream packages/markitdown-mcp/Dockerfile, a StatefulSet, a persistent /workdir volume, a Service, an HTTPS Ingress, and a Canvas App entry.
markitdown-mcp --http --host 0.0.0.0 --port 3001 command./workdir for trusted local files./mcp/ for Streamable HTTP, /sse for SSE, and /messages/ for SSE message posts.The upstream image enables its documented MarkItDown plugins and includes ffmpeg and exiftool. There are no database, S3, or login inputs. The server intentionally has no authentication; use a private Sealos workspace, an additional authenticated proxy, or an allowlisted network before sending sensitive documents.
| Component | Replicas | CPU limit | Memory limit | Storage |
|---|---|---|---|---|
| MarkItDown MCP | 1 | 100m | 256Mi | - |
| Work directory | 1 | - | - | 1Gi |
This is the initial personal low-load tier for protocol requests and small documents. Increase memory for large PDFs or conversions that include many embedded images.
MarkItDown and the official MCP package are licensed under the MIT License.
Sealos is an AI-assisted Cloud Operating System built on Kubernetes. Hosting the official MCP server on Sealos provides one-click provisioning, automatic HTTPS, persistent workspace storage, pay-as-you-go resources, and Canvas operations for a trusted agent environment.
/mcp/ or /sse./workdir across restarts.https://<host>/mcp/ for Streamable HTTP or https://<host>/sse for SSE.tools/list to verify the convert_to_markdown tool.For a Streamable HTTP client, use:
https://<your-markitdown-host>/mcp/
For SSE clients, use:
https://<your-markitdown-host>/sse
The server has no login page and no built-in authentication. Treat the endpoint as a privileged document-fetching service: convert_to_markdown can read files available to the container and fetch network resources allowed by its runtime.
Place trusted local files in /workdir and reference them with file:///workdir/<name>. Use Canvas resource cards to adjust CPU, memory, and storage. Keep one replica while using a ReadWriteOnce work directory; scale only after providing an external shared file strategy and an access-control layer.
Confirm the client uses /mcp/ for Streamable HTTP or /sse for SSE, and that the Sealos URL includes HTTPS. Check the MCP container logs and Service endpoints in Canvas.
Upload the file into /workdir and use a file:///workdir/... URI. The container cannot see paths on your laptop unless the bytes are uploaded into its workspace.
Increase the container memory tier and the Ingress read/send timeout from Canvas for large PDFs, embedded images, or audio conversions.
The upstream server does not support authentication. Restrict the public domain with network policy or an authenticated reverse proxy before using confidential documents, and keep the MCP endpoint limited to trusted clients.
This Sealos template is provided for Sealos users under the templates repository license. MarkItDown itself is licensed under the MIT License.
A shorter path from an app you want to an app you can run. Sealos brings deployment and ongoing operations into one place.
Start with a ready-made template. Review its configuration and launch from the Sealos console.
Run on managed infrastructure with built-in workload scheduling and recovery.
Give your application a public HTTPS endpoint with certificates managed for you.
Keep application data on persistent volumes across container restarts.
Adjust CPU, memory, and replicas from the console as your workload changes.
Sealos provisions the resources defined by your template and brings them together in your workspace.
Container workloads with configurable CPU, memory, and replicas.
An address for your app, with managed TLS certificates.
A database provisioned alongside your app when the template calls for one.
Storage for the files and data your application needs to keep.
Application settings and secrets configured in one place.
Inspect container logs and resource usage from the console.
Resources and their configuration vary by template. Review the deployment form for this app’s exact setup.
Explore more production-ready templates.
Build production-ready ai templates with Sealos.
Open Design is a local-first, open-source AI design workspace for generating prototypes, decks, images, videos, and design-system-driven artifacts.
Build production-ready ai templates with Sealos.
Host OpenClaw's AI agent gateway on Sealos with a browser control panel, token authentication, and persistent storage for configuration and workspaces.
Build production-ready ai templates with Sealos.
Open-source AI model gateway compatible with the OpenAI API, with multi-provider routing, quota management, channel orchestration, and an admin dashboard.