This MCP server gives AI agents structured read/write access to local Markdown files via selectors, without regex or string hacking.
Get a document's full heading outline (nodiom_tree).
Read raw Markdown at a section or node (nodiom_read).
Read all list items under a section as a JSON array (nodiom_read_list).
Replace a section's body while preserving its heading (nodiom_write).
Append content after the last item in a section (nodiom_append).
Delete a matched node or entire section (nodiom_delete).
Check existence and metadata like type, depth, and child count (nodiom_query).
Address headings, list items (li[0], li[-1], contains="..."), paragraphs (p[0]), tables (table[0]), and repeated headings via " > "-separated selectors.
Preserve YAML/TOML frontmatter and provide fuzzy suggestions when selectors don't match.
Provides structured read/write access to Markdown documents, allowing AI agents to read, write, append, and delete content using structural selectors without regex.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@nodiom-mcp Show the active tasks in /Users/me/projects/wiki.md"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server for nodiom — structured read/write access to Markdown documents for AI agents.
Expose nodiom's structural Markdown operations as Model Context Protocol tools. Any MCP-compatible agent (Claude Desktop, Claude Code, Cursor, or your own client) can read, write, append and delete content in Markdown files using structural selectors, without regex or string hacking.
Local files or shared state? This server works on Markdown files on your machine, for agents running there. When agents on different machines, from different vendors, or your teammates need the same documents, use Nodiom Cloud. It's the same engine, hosted, with shared boards, a human review step and attribution. Connect from Claude, ChatGPT, Cursor or VS Code by adding
https://api.nodiom.md/mcpand signing in.
npm install -g @synexiom-labs/nodiom-mcpOr use directly with npx (no install required):
npx @synexiom-labs/nodiom-mcpRelated MCP server: search-docs
Add to your claude_desktop_config.json:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"nodiom": {
"command": "npx",
"args": ["-y", "@synexiom-labs/nodiom-mcp"]
}
}
}claude mcp add -s user nodiom -- npx -y @synexiom-labs/nodiom-mcpThe -s user flag registers the server globally across all your projects. Without it, the server is only active when Claude Code's working directory matches the project where you ran the command.
Add to ~/.cursor/mcp.json (or .cursor/mcp.json in a project):
{
"mcpServers": {
"nodiom": {
"command": "npx",
"args": ["-y", "@synexiom-labs/nodiom-mcp"]
}
}
}{
"command": "npx",
"args": ["-y", "@synexiom-labs/nodiom-mcp"]
}The server exposes 7 tools, all operating on local Markdown files by absolute path.
nodiom_treeGet the structural outline of a document before reading or modifying it.
file: "/path/to/wiki.md"
→ Returns a nested JSON tree of all headingsnodiom_readRead the content at a structural location.
file: "/path/to/wiki.md"
selector: "# Project Aurora > ## Tasks > ### Active"
→ Returns the Markdown body of that section (the heading line itself isn't included)nodiom_read_listRead all list items under a section as a JSON array: each item's source, marker and nested lines included.
file: "/path/to/wiki.md"
selector: "# Project Aurora > ## Team"
→ ["- Alice Chen — Tech Lead", "- Bob Martinez — ML Engineer", ...]nodiom_writeReplace the content of a section (heading is preserved).
file: "/path/to/wiki.md"
selector: "# Project Aurora > ## Overview"
content: "Project Aurora is on track for Q3 delivery."
→ Replaces only the Overview body. Nothing else changes.Pointed at one list item (## Tasks > li[contains="staging"]), it replaces just that item. Text written without its own bullet keeps the item's marker and checkbox.
nodiom_appendAppend content after the last item in a section.
file: "/path/to/wiki.md"
selector: "# Project Aurora > ## Tasks > ### Active"
content: "- [ ] Deploy to staging"
→ Adds the new task at the end of Active. Existing tasks untouched.nodiom_deleteRemove a node or section.
file: "/path/to/wiki.md"
selector: "# Project Aurora > ## Tasks > ### Completed > li[0]"
→ Removes the first completed task.nodiom_queryCheck if a section exists and get its metadata.
file: "/path/to/wiki.md"
selector: "# Project Aurora > ## Tasks"
→ { "exists": true, "type": "heading", "depth": 2, "childCount": 3, "index": 4 }Selectors are " > "-separated paths of heading and element segments:
"# Project" → H1 section
"# Project > ## Tasks" → H2 under H1
"# Project > ## Tasks > ### Active" → H3 under H2 under H1
"## Tasks > li[0]" → First list item
"## Tasks > li[-1]" → Last list item
"## Notes > p[0]" → First paragraph
"## Arch > table[0]" → First table
"## Tasks > li[contains=\"auth\"]" → The list item containing "auth"
"## Notes[1]" → The second "## Notes", when the heading repeatsA content selector (li[contains="…"]) must match exactly one item, and keeps pointing at the same item when others are added or removed — prefer it whenever more than one writer touches a file. A heading that appears more than once must be addressed with an index (## Notes[0], ## Notes[1]); the error lists the options. YAML or TOML frontmatter is preserved and never treated as a heading.
When a selector doesn't match, the error includes fuzzy suggestions: "Did you mean '## Tasks'?"
Once the server is configured, you can instruct Claude naturally:
"Look at my project wiki at
/Users/me/projects/aurora/wiki.md. What are the active tasks? Add a new task: 'Write integration tests'. Then move the first completed task to a new '## Archive' section."
Claude will use nodiom_tree to understand the structure, nodiom_read_list to get the tasks, nodiom_append to add the new one, and nodiom_read + nodiom_delete + nodiom_append to move the completed task — all without loading the entire file as a string.
Package | Description |
The core library: use it directly in your Node.js code | |
| This package: an MCP server for AI agents, on local Markdown files |
Hosted: shared documents and boards for agents and people across machines, with sign-in from Claude, ChatGPT, Cursor and VS Code |
MIT — Synexiom Labs Inc.
Appends new content after the last item in a section. Use this to add a new task, log entry, or note without disturbing existing content. Content is added at the end of the matched section.
Inputs: file (required), content (required), selector (required).
Removes the node or section matched by the selector. Use "## Section > li[0]" to delete the first list item. Use "## Section" to delete an entire section including its heading. All other content is untouched.
Inputs: file (required), selector (required).
Returns structural metadata about a location: whether it exists, its type, depth, and child count. Use this to check if a section exists before reading or writing to it.
Inputs: file (required), selector (required).
Returns the Markdown content at a structural location in a file. Use a selector like "# Heading > ## Subheading" to address any section. Use "# H1 > ## H2 > li[0]" to read a specific list item. Returns the raw Markdown string of the matched content.
Inputs: file (required), selector (required).
Returns all list items under a selector as a JSON array of strings. Useful when you need to iterate over tasks, team members, or any bullet list.
Inputs: file (required), selector (required).
Returns the full structural outline of a Markdown document as a nested JSON tree of headings. Use this to understand the structure of a document before reading or modifying it.
Inputs: file (required).
Replaces the content at a structural location with new content. The heading itself is preserved — only the body content is replaced. All other sections in the document are untouched.
Inputs: file (required), content (required), selector (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 7 tools
Each tool targets a distinct operation on the Markdown structure: reading content, reading list items, writing, appending, deleting, querying metadata, and viewing the outline. There is no meaningful overlap that would cause an agent to select the wrong tool.
Tools consistently use the nodiom_ prefix with lowercase snake_case names. Most are verb-based (read, write, append, delete, query), though 'tree' is a noun rather than a verb and could have been 'get_tree' for full consistency.
Seven tools is a well-scoped set for structured Markdown editing. Each tool covers a distinct need without redundancy or bloat, and the count is appropriate for the server's focused purpose.
The toolset covers read, write, append, delete, and structural inspection well. A minor gap is the lack of an explicit create-section operation; a selector must presumably reference an existing heading, so creating wholly new sections may require workarounds.
Markdown workspace for AI agents: read, write, organize, and share markdown documents.
Portable AI memory shared across models and harnesses - plain markdown you own.
Markdown notes and whiteboards your AI agent can read and write.
Shared Markdown memory for your agents: docs, typed collections, files, skills and Dispatch.
