
ezmd
io.github.larpeyPublished by GitHub account larpeyv0.1.1Updated Oct 24, 2026
Convert files, URLs and text to LLM-ready Markdown with paging, provenance and warnings.
Overview
Local MCP server that converts files, URLs and pasted text into LLM-ready Markdown with paging, section ids and warnings.
- What you can do
- Convert an http(s) URL to Markdown, fetched through an SSRF guard and paginated to a token budget with provenance markers and warnings for skipped or truncated content. Convert a local file inside allowed directories, or a base64 data URL under 1 MB. Convert pasted text, HTML, Markdown, CSV or JSON. Page through results and check job status with a cursor-based job tool. List supported types, missing extras, limits and whether fetching is allowed.
- When to use it
- Useful when an assistant needs to read web pages, local documents or pasted content as clean Markdown for summarization or citation, especially when output must fit a token budget and cite section ids and pages.
- Requirements
- Runs locally as a stdio process, launched with uvx from the ezmd-mcp PyPI package, so Python and uv are needed. Optional settings: EZMD_MCP_ALLOWED_DIRS for readable directories (defaults to the working directory), EZMD_PROFILE for the default output profile, and EZMD_REMOTE plus EZMD_API_KEY to forward conversions to a remote ezmd instance. Network access is needed for URL conversion.
Installation
In SourceWeft
- Open ezmd in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
ezmd-mcp
An MCP server for ezmd: convert files, URLs and pasted text to LLM-ready Markdown from Claude Desktop, Claude Code, Cursor, VS Code and any other MCP client. Results are paginated to your token budget, carry section ids and page markers for citations, and report every warning (skipped pages, truncated tables, suspected prompt injection) instead of dropping content silently.
Tools
Every result has job_id, title, source, tokens_total, pages_total, page, next_cursor, warnings, frontmatter, content (page 1 adds sections). The default profile is agent (untrusted-content fence,
section ids); max_tokens defaults to 8000 and is capped at 50000. Resources ezmd://jobs/{id} and
ezmd://jobs/{id}/sidecar expose the full Markdown and sidecar JSON; the prompt
summarize_with_provenance asks the model to cite section ids and pages.
See docs/mcp.md for client configuration, security notes and every option. Apache-2.0.
Source: packages/mcp/README.md at commit b138ab9
Tools
0Version history
2- v0.1.1LatestOct 23, 2026
- v0.1.0Oct 20, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts