InfraNodus MCP Server
Officialby infranodus
README.md
# InfraNodus MCP Server
A Model Context Protocol (MCP) server that integrates InfraNodus knowledge graph and text network analysis capabilities into LLM workflows and AI assistants like Claude Desktop.
[](https://www.npmjs.com/package/infranodus-mcp-server)
Available on npm: [https://www.npmjs.com/package/infranodus-mcp-server](https://www.npmjs.com/package/infranodus-mcp-server)
## Overview
InfraNodus MCP Server enables LLM workflows and AI assistants to analyze text using advanced network science algorithms, generate knowledge graphs, detect content gaps, and identify key topics and concepts. It transforms unstructured text into structured insights using graph theory and network analysis.
## Features
### You Can Use It To
- Connect your existing InfraNodus knowledge graphs to your LLM workflows and AI chats
- Identify the main topical clusters in discourse without missing the important nuances (works better than standard LLM workflows)
- Identify the content gaps in any discourse (helpful for content creation and research)
- Generate new knowledge graphs from any text and use them to augment your LLM responses
- Save and retrieve entities and relations from memory using the knowledge graphs
### Available Tools
1. **generate_knowledge_graph**
- Convert any text into a visual knowledge graph
- Extract topics, concepts, and their relationships
- Identify structural patterns and clusters
- Apply AI-powered topic naming
- Perform entity detection for cleaner graphs
2. **analyze_existing_graph_by_name**
- Retrieve and analyze existing graphs from your InfraNodus account
- Access previously saved analyses
- Export graph data with full statistics
3. **analyze_text**
- Analyze a text, URL, or YouTube transcript
- Extract and analyze a graph from text or URL; provide either text or url
- Get topics, clusters, statements, graph structure, and AI summary as requested
4. **analyze_text_signature**
- Structural and rhythmic signature of a text, URL, or YouTube transcript
- Every measure with a plain-language reading, a named profile (essay, digest, anthology...), and a hedged AI-likeness estimate
- Use to compare writing styles, characterise an author or genre, or flag flat, uniformly paced text
5. **generate_content_gaps**
- Detect missing connections in discourse
- Identify underexplored topics
- Generate research questions
- Suggest content development opportunities
6. **generate_topical_clusters**
- Generate topics and clusters of keywords from text using knowledge graph analysis
- Make sure to beyond genetic insights and detect smaller topics
- Use the topical clusters to establish topical authority for SEO
- Returns AI-generated overviews of the topical clusters (`topicalClusterSummaries`), summarizing the discourse each cluster represents — useful for SEO-optimized content creation. Enabled by default; set `generateTopicalSummaries: false` to increase processing speed or if the summary request fails
7. **generate_contextual_hint**
- Generate a topical overview of a text and provide insights for LLMs to generate better responses
- Use it to get a high-level understanding of a text
- Use it to augment prompts in your LLM workflows and AI assistants
8. **generate_research_questions**
- Generate research questions that bridge content gaps from text, URL, or an existing InfraNodus graph
- Use them as prompts in your LLM models and AI workflows
- Use any AI model (included in InfraNodus API)
- `optimize` chooses the clusters the questions bridge: `optimize` (default) the insight clusters picked from the structure of the graph, `gap` the widest content gap, `develop` the main topics, `latent` the under-developed ones
- Returns the insight clusters used as `insightClusters`
9. **generate_research_ideas**
- Generate innovative research ideas based on content gaps identified in the text
- Get actionable ideas to improve the text and develop the discourse
- Use any AI model (included in InfraNodus API)
- `optimize` chooses the clusters the ideas bridge, as in `generate_research_questions`; the default is the insight clusters, returned as `insightClusters`
10. **optimize_text_structure**
- Analyze the level of bias and coherence in text using knowledge graph analysis
- If the text is too biased, develop the represented topics to balance the discourse
- If the text is focused or diversified, develop the content gaps to deepen the analysis
- If the text is dispersed, focus the most common gap topics to improve coherence
- Choose response type: response, idea, question, or transcend
11. **optimize_reasoning**
- Applies the same bias/coherence analysis to the model's own reasoning trace or chat with the user (pass it as `text`)
- Detects whether the reasoning is biased, focused, diversified, or dispersed
- Steers the reasoning toward optimal diversity and coherence at the same time
- If too biased, develop the under-represented topics; if focused or diversified, bridge the content gaps; if dispersed, focus the most common gap topics
- Returns a structural diagnosis (diversity stats, topical clusters, gaps) plus suggestions for how to continue thinking
12. **generate_responses_from_graph**
- Generate responses based on an existing InfraNodus graph
- Integrate them into your LLM workflows and AI assistants
- Use any AI model (included in InfraNodus API)
- Use any prompt
13. **develop_conceptual_bridges**
- Analyze text and develop latent ideas based on concepts that connect this text to a broader discourse
- Discover hidden themes and patterns that link your text to wider contexts
- Use any AI model (included in InfraNodus API)
- Generate insights that help develop the discourse
14. **develop_latent_topics**
- Analyze text and extract underdeveloped topics with ideas on how to develop them
- Identify topics that need more attention and elaboration
- Use any AI model (included in InfraNodus API)
- Get actionable suggestions for content expansion
15. **develop_text_tool**
- Comprehensive text analysis combining content gap ideas, latent topics, and conceptual bridges
- Executes multiple analyses in sequence with progress tracking
- Generates research ideas based on content gaps
- Identifies latent topics and conceptual bridges to develop
- Finds content gaps for deeper exploration
16. **create_knowledge_graph**
- Create a knowledge graph in InfraNodus from text and provide a link to it
- Use it to create a knowledge graph in InfraNodus from text
- Also accepts `statements` (discrete items) with parallel `categories` and `timestamps`. Categories are stored as metadata on each statement (filter, group, delete, or relabel by them) and stay out of the graph unless `categoriesAsNodes: true`, which turns each label into a `[[label]]` node — meant for large knowledge bases and Obsidian vaults where the connections between pages are the point. The parent modes of `wikilinksMode` (`obsidianStyle`, `parentAndConcepts`) always make the parent page a node
17. **generate_ontology_graph**
- Use AI to generate a reasoning ontology graph (entities and the relations between them) from one of three sources: a `prompt` (a topic — e.g. "build an ontology on AI attention mechanisms"), a `text` (a long document or a structural digest of a project, chunked server-side), or a `sourceGraphName` (an existing graph — e.g. a fully ingested repo or corpus — whose statements are read back, chunked, and condensed into an ontology)
- `ontologyMode: 'codebase'` frames the extraction around modules, functions, data stores, services, and the concepts they implement; `ontologyMode: 'procedural'` writes a **digest** instead — prose statements on how the project works with `[[wikilinks]]`, each typed as `[principles]`, `[rules]`, `[procedures]`, `[handoffs]`, `[main_ideas]` or `[gaps]` (saved as the statement's category), from an already-uploaded docs/structure graph; save it as `repo-<project>-digest` for `optimize_knowledge_base`; `chunkSize` (default 12000 chars) controls granularity; every chunk appends to the same graph and the response reports `chunksProcessed` / `chunksTotal`
- Saved as a persistent InfraNodus graph by default and a link is returned; set `saveGraph: false` if the user asks not to save, or when you only need a one-off AI ontology overview of a topic for the current context that won't be reused later (the generated statements are returned directly without persisting)
- `modelToUse` defaults to `claude-opus-5-5` for richer ontologies (also `claude-fable-5-1`, `gpt-6.1-terra`); pick `-mini`/`-lite` variants (or `gpt-4o-mini`) for faster, cheaper generation
- Returns the compact graph structure (`knowledgeGraph`) and analytics (main topical clusters, content gaps, top influential nodes, top relations, statistics) by default. Set `includeGraph: false` to save context space when only the ontology statements or insights are needed. Set `includeAnalytics: false` if you just need the raw ontology without graph-derived insights — keep it on whenever you want to understand the structure, gaps, or key concepts
18. **analyze_llm_results**
- Ask an LLM to describe a topic and turn its response into a knowledge graph that reveals how the model frames it — main concepts, clusters, content gaps, and the relations between them
- Use it to probe model bias, surface the implicit structure of an LLM's view on a subject, or compare how different models describe the same topic
- `modelToUse` defaults to `claude-opus-5-5`; pick the model you actually want to study
- `modifyAnalyzedText` controls how the LLM output is parsed: `'detectEntities'` (default — mixed entities + words), `'extractEntitiesOnly'` (entity-only graph), or `'none'` (plain co-occurrence)
- Saves the graph by default; set `saveGraph: false` for a one-off probe. Returns analytics by default and omits the raw graph (`includeGraph: false`) to keep responses compact — enable `includeGraph` when you also need nodes/edges
19. **overlap_between_texts**
- Create knowledge graphs from two or more texts and find the overlap (similarities) between them
- Use it to find similar topics and keywords across different texts
20. **merged_graph_from_texts**
- Build a graph of all the texts and URLs provided, providing topical clusters and gaps present in the merged graph generated from all the texts
- Use it to combine multiple sources into one graph and see clusters and content gaps across the merged content
21. **difference_between_texts**
- Compare knowledge graphs from two or more texts and find what's not present in the first graph that's present in the others
- Use it to find how one text can be enriched with the others
22. **analyze_google_search_results**
- Generate a graph with keywords and topics for Google search results for a certain query
- Use it to understand the current informational supply (what people find)
23. **analyze_youtube_results**
- Generate a graph with keywords and topics from YouTube results for a query, channel, or playlist
- Choose what to pull via `searchMode`: `search` (video metadata for a search term), `comments` (comments on a video), `channel` (a channel's videos — pass a username, URL, or @handle), `playlist` (a playlist's videos — pass a playlist ID or a URL with `list=`), `subtitles` / `subtitlesChannel` / `subtitlesPlaylist` (transcribed subtitles of a video / channel / playlist), or `searchVideos` (analyzes the content of the videos found — limit hard-capped to 20)
- Control results with `limit` (default 100, max 2000), `sortBy` (`Popular` / `Oldest` / `Latest`), `excludeDescriptions`, `importLanguage`, and `importRegion`
- Use it to understand the topics, clusters, and content gaps in the discourse around a video, channel, playlist, or search term on YouTube
24. **analyze_related_search_queries**
- Generate a graph from the search queries suggested by Google for a certain query
- Use it to understand the current informational demand (what people are looking for)
25. **search_queries_vs_search_results**
- Generate a graph of keyword combinations and topics people tend to search for that do not readily appear in the search results for the same queries
- Use it to understand what people search for but don't yet find
26. **generate_seo_report**
- Analyze content for SEO optimization by comparing it with Google search results and search queries
- Identify content gaps and opportunities for better search visibility
- Get comprehensive analysis of what's in search results but not in your text
- Discover what people search for but don't find in current results
27. **memory_add_relations** (Save to InfraNodus Memory)
- Save statements to a named persistent memory graph and get back its name and URL
- For "remember this", "save this to memory", or ideas, plans, and decisions worth keeping
- With `graphName`: writes immediately to that graph, no opt-in, no dry run; reuse the name to append
- With `scope` instead: `topic` (+ `topic`, graph `memory-<topic>`) for an outside subject, client or case, or `personal` (graph `personal-<user>`) for your own preferences and setup. These must be enabled once with `enable_memory`, and each batch is a dry run until approved, like project learnings
- Mark entities with [[wikilinks]] (at least two per statement); pass categories and timestamps for later filtering
28. **memory_get_relations** (Recall from InfraNodus Memory)
- Recall what was previously saved to memory about a concept or entity
- For "what do I have on X", "did we decide X", "check memory"
- Query one memory graph by name or search across all memory graphs
- Returns the statements and relationships that mention the entity
- With `scope` (`topic` + `topic`, or `personal`): reads that kept memory by entity, by `prompt` (GraphRAG), or as an overview
29. **enable_memory**
- Create your topic memory (`memory-<topic>`) or personal memory (`personal-<user>`) graph so the memory tools can use it with `scope`
- Called only when you explicitly ask to keep that kind of memory; idempotent
30. **retrieve_from_knowledge_base**
- Retrieve context from an existing InfraNodus knowledge graph using GraphRAG
- Query your knowledge base with a natural language prompt to get relevant statements
- Include graph summaries for quick overviews of the knowledge structure
- Optionally retrieve the full graph, statements, or extended analysis
- Ideal for augmenting LLM responses with domain-specific knowledge
31. **search**
- Search through existing InfraNodus graphs
- Also use it to search through the public graphs of a specific user
- Compatible with ChatGPT Deep Research mode via Developer Mode > Connectors
32. **fetch**
- Fetch a specific search result for a graph
- Can be used in ChatGPT Deep Research mode via Developer Mode > Connectors
33. **get_graph_around_term**
- The graph around a term across your graphs: every statement that contains the term — or all of several terms (`matchAll`, the default) or any of them — in all your graphs or the named ones, built into one graph with the terms themselves left out, so the topics, gaps and relations describe the context the terms sit in (the app's search view with the searched term hidden)
- Returns which graphs the statements came from with a count each, the statistics, topical clusters, content gaps, main concepts and top relations, the compact graph (`includeGraph`, on by default) and the statements with their graph of origin (`includeStatements`, off by default); `removeTerms: false` keeps the terms in
- Your own graphs only; `search` says where a term occurs, `get_node_neighborhood` walks one graph from one concept
34. **list_graphs**
- List the graphs (contexts) in your InfraNodus account, with optional filters by name, type, date, language, or favorite status
- Use it to discover which graphs exist before analyzing them with `analyze_existing_graph_by_name`, searching them, or writing to them
35. **enable_project_learnings**
- Create the opt-in, per-project, append-only learnings graph (`learn-<project>`) in your account — the place where the assistant saves what it learned about operating in a project
- Called only when you explicitly ask to start saving learnings for a project; idempotent, so calling it again just returns the existing graph
- The assistant tells you first what will be stored (project knowledge only, never anything about you), where (a private graph you can delete at any time), and that batches are shown before saving
36. **add_project_learnings**
- Save learnings about a project — where things live, traps, conventions, decisions, workflows, and a self-assessment of what worked well and what should be done differently next time — as statements with a `type` category each, for later sessions on any client
- Refuses (without error) when the project has not been enabled and never creates the graph itself
- Dry run by default: returns what would be written, marking near-duplicates as `reinforced`; writes only with `confirm: true` — or in the same call when your client supports MCP elicitation and you approve the form
- Rejects statements with secret-like content server-side (indices only, never the content)
37. **get_project_learnings**
- Retrieve learnings for a project: by `prompt` (GraphRAG — most relevant statements plus an overview of what is known), by `entity` (a file path, module, or concept), or a structural overview with neither
- Call with no `project` to list the projects that have learnings in your account
- Returns `enabled: false` with an empty list when a project has no learnings graph — not an error
_Also available as the `save-learnings` prompt in clients that expose MCP prompts. Set `INFRANODUS_LEARNINGS=0` to remove these three tools from the server entirely (see [Project learnings](#project-learnings))._
38. **optimize_knowledge_base**
- Structural feedback on a whole code base, document vault, or body of procedural knowledge (rules, frameworks, principles) from its digest graph (`repo-<project>-digest` / `vault-<project>-digest`) — or from a digest you write and pass as text, saved with `saveAs`
- Diagnoses the structure (biased / focused / diversified / dispersed) and reads it for the chosen `focus` (`codebase`, `vault`, `procedural`, `general`): what dominates, which areas are under-developed (with latent-topic ideas), which clusters never connect — missing integrations, missing bridge notes, missing hand-offs between frameworks — plus AI suggestions for what to develop next
- `optimize` chooses what the suggestions work on: `optimize` (default) the insight clusters, which are returned as `insightClusters` and stepped through with `gapDepth`; `develop` all the main topics, followed by a latent-topics pass; `latent` the under-developed ones. `includeLatent: true` adds the latent pass to `optimize`
- `compareWith` up to two other layers of the same project (e.g. digest vs structure map, docs vs code) and get what each has that the other lacks: rules without code, code without documentation, features described but not built
- Two ways to get the digest: the assistant reads the project and writes it (the infranodus skill's **digest** build mode, or `text` + `saveAs` here), or `generate_ontology_graph` with `ontologyMode: 'procedural'` writes it server-side from an uploaded docs/structure graph
- Pairs with the infranodus skill's build modes: full ingestion, digest, structure map, ontology
39. **submit_workflow_feedback**
- Internal telemetry: after a workflow of InfraNodus calls, the assistant reports what it actually did with the output — how much it used, whether it contained anything new, whether it was the right tool, how many calls it took, concrete defects — as observations rather than a score; a rating is derived from them server-side
- Autonomous: it never asks you anything and never surfaces in the conversation
- The report contains short paraphrases of what you were working on (capped at 200 characters), the same policy as the prompts InfraNodus already logs; `MCPCAT_ANONYMOUS=1` skips the per-user log entirely
- Set `INFRANODUS_FEEDBACK=0` to remove the tool (and the one-line nudge appended to workflow-ending results) from the server
39. **get_more_tools**
- Added by the [mcpcat](https://mcpcat.io) analytics library: the assistant calls it when a task would benefit from a capability none of the tools above provides, describing what it needed (not offered on the `/openai` endpoint used by the ChatGPT / Codex plugin)
- The server has no hidden tools, so the call only records the request as feedback for future tool development; the assistant does not need to ask you before calling it
- Its logs follow the same privacy rules as the usage analytics below (`MCPCAT_ANONYMOUS=1` keeps them unlinked from your account)
40. **delete_statements**
- Delete statements from a graph in your own account by a filter the server resolves — exactly one of `categories` (everything uploaded under a source label: a file path, a page name, a `[[label]]` parent), `statements` (exact text), `query` (substring or `/regex/`), `before`/`after` (ISO 8601 window), `deleteAll` (empties the graph but keeps its name, URL, and settings), or `statementIds`
- Dry run by default: returns the matched count, a sample of the statements, and a per-category breakdown; nothing is removed until the same filter is sent again with `confirm: true` — or, on clients with MCP elicitation, until you accept the form in the same call. A filter that matches nothing returns `deleted: 0` without asking
- The per-source replace path: `delete_statements` with the source's category, then `create_knowledge_graph` to the same graph name. `deleteAll` then `create_knowledge_graph` rebuilds a graph in place
- Irreversible, own-account only (no `userName`), never creates a graph, and the assistant is instructed never to call it on its own initiative. Set `INFRANODUS_DELETE=0` to remove the tool from the server
41. **update_statements**
- Edit statements of a graph in your own account **in place** — content, categories, or timestamp — keeping each statement's id, date, and position (unlike deleting and re-creating it)
- Two modes: `edits` rewrites specific statements, each named by its exact current text (`match`, e.g. from `analyze_existing_graph_by_name` with `includeStatements`) or `statementId`; or one selector (the same as `delete_statements`, with `all` instead of `deleteAll`) plus `set` (`addCategories`, `removeCategories`, `categories`, `timestamp`) and/or `replace` (`{ pattern, with }`, substring or `/regex/flags`) for bulk relabelling or a find-and-replace across the graph — renaming a `[[concept]]` or a source path everywhere
- Dry run by default: returns the matched count and the before → after of every change; nothing is written until the same arguments are sent again with `confirm: true` — or, on clients with MCP elicitation, until you accept the form in the same call. A request that matches nothing returns `updated: 0` without asking
- New content is capped at 1000 characters; for longer text use `delete_statements` then `create_knowledge_graph`. Irreversible (the old text survives only in the dry-run output), own-account only, never creates a graph, and the assistant is instructed never to call it on its own initiative. `INFRANODUS_DELETE=0` removes it together with `delete_statements`
42. **delete_graph**
- Delete one graph from your own account entirely — its statements, revisions, settings, and URL; the name becomes free again. `delete_statements` with `deleteAll` is the alternative that keeps the graph for a rebuild in place
- Dry run by default: confirms the graph exists and reports its URL and whether it is a live graph; nothing is removed until the same `graphName` is sent again with `confirm: true`. Clients that support elicitation ask you directly and delete in the same call
- Irreversible, own-account only (no `userName`; the graph is resolved from your API key's own graph list and the backend scopes the delete to the same user), and the assistant is instructed never to call it on its own initiative. `INFRANODUS_DELETE=0` removes it together with the other two mutation tools
_More capabilites coming soon!_
### Key Capabilities
- **Topic Modeling**: Automatic clustering and categorization of concepts
- **Content Gap Detection**: Find missing links between concept clusters
- **Entity Recognition**: Clean extraction of names, places, and organizations
- **AI Enhancement**: Optional AI-powered topic naming and analysis
- **Structural Analysis**: Identify influential nodes and community structures
- **Network Structure Statistics**: Modularity, centrality, betweenness, and other graph metrics
- **Fractal Variability**: DFA scaling exponents (alpha) of the discourse at three path levels (statements, words, and n-grams of 4 consecutive words), by step length and radial distance, plus the sentence length level (words per sentence, in order) and the influence level (betweenness, degree, and betweenness rank of the current word; read the rank series), with multifractal spectra — returned as `fractal_variability` next to `diversity_stats` whenever the graph statistics are included
- **Degree Distribution**: `degree_distribution` on every graph — nodes, edges, a `[degree, count]` histogram (first 40 rows unless `fullGraph`), gini of degree, and a power-law tail exponent over the top nodes (null under 50 nodes) — descriptors of how concentrated connectivity is in the whole co-occurrence network before the node cap; never a scale-free verdict
- **Knowledge Graph Memory**: Save and retrieve knowledge graph memories and analyze them to retrieve key nodes, clusters, and connectors
## Knowledge Graph Memory Use Advice
InfraNodus represents any text as a network graph in order to identify the main clusters of ideas and gaps between them. This helps generate advanced insights based on the text's structure. The network is effectively a knowledge graph that can also be used to retrieve complex ontological relations between different entities and concepts. This process is automated in InfraNodus using the `search` and `fetch` tools along with the other tools that analyze the underlying network.
However, you can also easily use InfraNodus as a more traditional memory server to save and retrieve relations. We use [[wikilinks]] to highlight entities in your text to make your content and graphs compatible with markup syntax and PKM tools such as Obsidian. By default, InfraNodus will generate the name of the memory graph for you based on the context of the conversation. However, you can modify this default behavior by adding a **system prompt** or **project instruction** into your LLM client.
Specifically you can specify to always use a speciic knowlege graph for memories to store everything in one place:
```
Save all memories in the `my-memories` graph in InfraNodus.
```
Or you can ask InfraNodus to only save certain entities, e.g. for building social networks:
```
When generating entities, only extract people, companies, and organizations. Ignore everything else.
```
## Installation
The easiest and the fastest way to launch the InfraNodus MCP server is to either use our server URL `https://mcp.infranodus.com` for the remote / web applications or to add a manual configuration to your LLM apps if you're running them locally.
You can also install the server locally, so you have more control over it. In this case, you can also edit the source files and even create your tools based on the [InfraNodus API](https://infranodus.com/api).
Below we describe the two different ways to set up your InfraNodus MCP server.
### 1. Easiest Setup: InfraNodus MCP Server (via HTTP/SSE)
0. **Prerequisites**
- Create an account on [InfraNodus](https://infranodus.com) if you don't have it already and get your [InfraNodus API Key](https://infranodus.com/api-access). We offer 14-day free trials.
1. **Get the URL**
- We currently use the following URL for our MCP server deployed in our infrastructure:
```bash
https://mcp.infranodus.com
```
2. **Add the MCP server URL to the Client Tool Where You Want to Use InfraNodus**
- Once you add the URL above to your tool, it will automatically prompt you to authenticate using OAuth in order to be able to access the InfraNodus MCP hosted on it.
4. **Using InfraNodus Tools in Your Calls**
- To use InfraNodus, see the tools available and simply call them through the chat interface (e.g. "show me the graphs where I talk about this topic" or "get the content gaps from the document I uploaded")
- If your client is not using InfraNodus for some actions, add the instruction to use InfraNodus explicitly.
### 2. Manual Setup: via NPX
You can deploy the InfraNodus server manually via `npx` — a package that allows to execute local and remote Node.Js packages on your computer.
The InfraNodus MCP server is available as an npm package at [https://www.npmjs.com/package/infranodus-mcp-server](https://www.npmjs.com/package/infranodus-mcp-server) from where you can launch it remotely on your local computer with npx. It will expose its tools to the MCP client that will be using this command to launch the server
#### For Claude Desktop / Cursor IDE:
Just add this in your Claude's configuration file (Settings > Developer > Edit Config), inside the `"mcpServers"` object where the different servers are listed:
```json
{
"mcpServers": {
"infranodus": {
"command": "npx",
"args": ["-y", "infranodus-mcp-server"],
"env": {
"INFRANODUS_API_KEY": "YOUR_INFRANODUS_API_KEY"
}
}
}
}
```
#### For Claude Code
To connect the InfraNodus MCP server to your Claude code, you can use this command. Make sure to provide the correct InfraNodus API key for your account:
```bash
claude mcp add infranodus -s user \
-- env INFRANODUS_API_KEY=YOUR_INRANODUS_KEY \
npx -y infranodus-mcp-server
```
### 3. Manual Setup: Launching MCP as a Local Server (for inspection & development)
0. **Prerequisites**
- Node.js 18+ installed
- InfraNodus API key (get yours at [https://infranodus.com/api-access](https://infranodus.com/api-access))
1. **Clone and build the server:**
```bash
git clone https://github.com/yourusername/mcp-server-infranodus.git
cd mcp-server-infranodus
npm install
npm run build:inspect
npm run inspect
```
Note that `build:inspect` will generate the `dist/index.js` file which you will then use in your server setup. The standard `npm run build` command will only build a Smithery file.
2. **Set up your API key:**
Create a `.env` file in the project root:
```
INFRANODUS_API_KEY=your-api-key-here
```
3. **Inspect the MCP:**
```bash
npm run inspect
```
### Claude Desktop Configuration (macOS)
1. Open your Claude Desktop configuration file:
```bash
open ~/Library/Application\ Support/Claude/claude_desktop_config.json
```
2. Add the InfraNodus server configuration:
a. remote launch via `npx`:
```json
{
"mcpServers": {
"infranodus": {
"command": "npx",
"args": ["-y", "infranodus-mcp-server"],
"env": {
"INFRANODUS_API_KEY": "YOUR_INFRANODUS_API_KEY"
}
}
}
}
```
b. launch this repo with `node`, specify the absolute path to the repo + `/dist/index.js`:
```json
{
"mcpServers": {
"infranodus": {
"command": "node",
"args": ["/absolute/path/to/mcp-server-infranodus/dist/index.js"],
"env": {
"INFRANODUS_API_KEY": "your-api-key-here"
}
}
}
}
```
**Note:** you can leave the `INFRANODUS_API_KEY` empty in which case you can make 70 free requests after which you will hit quota and will need to add your API key.
3. Restart Claude Desktop to load the new server.
### Claude Desktop Configuration (Windows)
1. Open your Claude Desktop configuration file:
```
%APPDATA%\Claude\claude_desktop_config.json
```
2. Add the InfraNodus server configuration:
a. remote launch via `npx`:
```json
{
"mcpServers": {
"infranodus": {
"command": "npx",
"args": ["-y", "infranodus-mcp-server"],
"env": {
"INFRANODUS_API_KEY": "YOUR_INFRANODUS_API_KEY"
}
}
}
}
```
b. launch this repo with `node`:
```json
{
"mcpServers": {
"infranodus": {
"command": "node",
"args": ["C:\\path\\to\\mcp-server-infranodus\\dist\\index.js"],
"env": {
"INFRANODUS_API_KEY": "your-api-key-here"
}
}
}
}
```
3. Restart Claude Desktop.
### Cursor Configuration
### Other MCP-Compatible Applications
For other applications supporting MCP, use the following command to start the server via npx:
```bash
INFRANODUS_API_KEY=your-api-key npx -y infranodus-mcp-server
```
or locally
```bash
INFRANODUS_API_KEY=your-api-key node /path/to/mcp-server-infranodus/dist/index.js
```
The server communicates via stdio, so configure your application to run this command and communicate through standard input/output.
### Legacy Setup via Smithery
InfraNodus server is also available through Smithery: a repository of MCP servers that has an easy-to-follow installation process for most LLM clients. You will need a separate accout at Smithery though.
- Create an account on [Smithery.Ai](https://smithery.ai/) (it's free and you can use your Google or GitHub login)
- Then go to the [Smithery InfraNodus Server](https://smithery.ai/server/@infranodus/mcp-server-infranodus), click "Configure" at the top right, and add your InfraNodus API key there.
- Go to [Smithery InfraNodus Server](https://smithery.ai/server/@infranodus/mcp-server-infranodus) and get the URL link from Smithery [https://server.smithery.ai/@infranodus/mcp-server-infranodus/mcp](https://server.smithery.ai/@infranodus/mcp-server-infranodus/mcp) for the server or use one of their automatic setup tools for Claude or Cursor.
- You may need to get your separate Smithery API key and Smithery proile link to make this work.
##### For Cursor:
```json
// e.g. Cursor will access directly the server via Smithery
"mcpServers": {
"mcp-server-infranodus": {
"type": "http",
"url": "https://server.smithery.ai/@infranodus/mcp-server-infranodus/mcp?api_key=YOUR_SMITHERY_KEY&profile=YOUR_SMITHERY_PROFILE",
"headers": {}
}
}
```
#### For Claude:
```json
// Claude uses a slightly different implementation
// Fot this, it launches the MCP server on your local machine
"mcpServers": {
"mcp-server-infranodus": {
"command": "npx",
"args": [
"-y",
"@smithery/cli@latest",
"run",
"@infranodus/mcp-server-infranodus",
"--key",
"YOUR_SMITHERY_KEY",
"--profile",
"YOUR_SMITHERY_PROFILE"
]
}
}
```
**Note**, in both cases, you'll automatically get the `YOUR_SMITHERY_KEY` and `YOUR_SMITHERY_PROFILE` values from Smithery when you copy the URL with credentials. These are not your InfraNodus API keys. You can use the InfraNodus API server without the API for the first 70 calls. Then you can add it to your Smithery profile and it will automatically connect to your account using the link above.
## Usage Examples
Once installed, you can ask Claude to:
- "Use InfraNodus to analyze this text and show me the main topics"
- "Generate a knowledge graph from this document"
- "Find content gaps in this article"
- "Retrieve my existing graph called 'Research Notes' from InfraNodus"
- "What are the structural gaps in this text?"
- "Identify the most influential concepts in this content"
## Development
### Running in Development Mode
```bash
npm run dev
```
### Using the MCP Inspector
Test the server with the MCP Inspector:
```bash
npm run build:inspect
npm run inspect
```
### Building from Source
```bash
npm run build
```
### Watching for Changes
```bash
npm run watch
```
### Running the Tests
```bash
npm test
```
Offline tests, no network: they compile the server and drive every tool handler through an intercepted `fetch`. `test/statsSurface.test.mjs` covers the network-statistics surface of all tools that return `diversity_stats` and `fractal_variability` (request flags and response shape). CI runs this on every pull request and on every push to `master`.
```bash
INFRANODUS_API_KEY=... npm run test:live
```
Live tests against the real API (`test/live/`): the same tools, checked for the actual data surface, including a computed multifractal spectrum on a long text. Set `INFRANODUS_API_BASE` to point at another API root, and `LIVE_SKIP_AI=1` to skip the tools that call an LLM. The saving tools create one throwaway graph and delete it afterwards. CI runs this after a merge into `master` and on demand from the Actions tab (workflow "Tests" → "Run workflow"); it needs the `INFRANODUS_API_KEY` repository secret and skips itself when the secret is absent.
## API Documentation
### generate_knowledge_graph
Analyzes text and generates a knowledge graph.
**Parameters:**
- `text` (string): The text to analyze. Provide this, `url`, or `statements`
- `statements` (string[]): Already-separated statements, with optional parallel `categories` (string[][]) and `timestamps` (ISO 8601 strings)
- `categoriesAsNodes` (boolean, default false): Show category labels as `[[label]]` nodes. Only for large knowledge bases and Obsidian vaults where the connections between pages matter; otherwise the labels stay metadata
- `wikilinksMode` (string): How `[[wikilinks]]` become nodes — "default", "wikilinksOnly", "obsidianStyle", "parentAndConcepts", or "plainText"
- `includeStatements` (boolean): Include original statements in response
- `modifyAnalyzedText` (string): How nodes are made — "none" (words), "detectEntities" (an algorithm marks names, mixed with words), "extractEntitiesOnly" (only the names an algorithm finds; for plain text without `[[wikilinks]]`), or "aiKnowledgeGraph" (an AI model writes `[[entity]]` statements about a topic or a text, and only those entities become nodes; slower, uses AI credits). Text already written as `[[entity]]` statements goes in with `wikilinksMode: "wikilinksOnly"`
- `modelToUse` (string), `instructions` (string): Only with "aiKnowledgeGraph" — the AI model, and the user's own instructions (which entities, scope); they cannot change the output format
### analyze_existing_graph_by_name
Retrieves and analyzes an existing graph from your InfraNodus account.
**Parameters:**
- `graphName` (string, required): Name of the existing graph
- `includeStatements` (boolean): Include statements in response
- `includeGraphSummary` (boolean): Include graph summary
### analyze_text
Analyze a text, URL, or YouTube transcript. Extract and analyze a graph from text or URL; provide either text or url.
**Parameters:**
- `text` (string, optional): Text to analyze. Provide either this or url.
- `url` (string, optional): URL to fetch content from (e.g. webpage or YouTube transcript). Provide either this or text.
- `includeStatements` (boolean): Include processed statements in response
- `includeGraph` (boolean): Include full graph structure in response
- `addNodesAndEdges` (boolean): Include nodes and edges in response
- `multifractal` (boolean): Also compute the multifractal spectrum inside `fractal_variability` (off by default, ~0.5 s per graph; only returned together with the graph statistics)
- `includeGraphSummary` (boolean): Include AI-generated graph summary for RAG prompt augmentation
- `modifyAnalyzedText` (string): Entity detection — "none", "detectEntities", or "extractEntitiesOnly"
### analyze_text_signature
Returns the structural and rhythmic signature of one text: how the discourse moves through its concept network, how its topics are distributed, and the sentence-length rhythm of the source. Every number comes with a one-sentence reading. Not a detector on its own.
**Parameters:**
- `text` / `url` / `statements`: the content, as in `analyze_text`
- `multifractal` (boolean, default true): compute the multifractal spectrum of the topic path (adds ~0.5–1 s)
- `maxNodes` (number): concepts kept in the graph (default 150); raise for long texts so the word path is complete
- `modifyAnalyzedText` (string): entity detection, as in `analyze_text`
**Response:**
- `signature`: a named profile from structure x rhythm (biased / focused / diversified / dispersed x alternating / random / fractal / blocks), e.g. `essay`, `report`, `narrative`, `digest`, `anthology`, `notes`, with a one-line summary
- `structure`: `modularity`, `diversity_stats` and their readings
- `rhythm`: `fractal_variability` (statements, words, ngrams x step length, radial distance; sentence length (words per sentence, in order) x byWords) and per-level readings of the DFA exponents, the short-vs-long-scale contrast, the multifractal label, and a confidence from the series length
- `amplitude`: measured on the same paths, in units of the graph radius — mean / median / p95 step, step-length CV (burstiness of movement), lag-1 autocorrelation, share of steps crossing a topic cluster and the regularity of those crossings, radial distance stats
- `sentence_rhythm`: sentence count, mean length, length CV (classic burstiness), lag-1 autocorrelation, share of short and long sentences, computed on prose-like statements only (fragments such as headings and menu items are skipped and counted); plus `scaling`, a summary of the backend's DFA on the sentence-length series (`fractal_variability.sentenceLength.byWords`, null under 64 sentences) with its own reading
- `ai_likeness`: `verdict` (leans generated / leans human / inconclusive), `score` (−1 human-like to +1 generated-like), `confidence`, the weighted `evidence` behind it, and `caveats`. It is a stylistic tendency: on a 50-page check against a commercial detector it caught uniformly paced template text but read AI-drafted pages that had been edited as human.
### generate_content_gaps
Identifies content gaps and missing connections in text.
**Parameters:**
- `text` (string, required): The text to analyze for gaps
## Progress Notifications
For long-running operations (like SEO analysis), the MCP server supports **real-time progress notifications** that provide intermediary feedback to AI agents. This allows agents to:
- Track the progress of multi-step operations
- Display status messages to users
- Understand what's happening during lengthy analyses
### Implementation
The server implements MCP progress notifications using:
1. **ToolHandlerContext**: All tool handlers can receive an optional context parameter containing the server instance and progress token
2. **ProgressReporter**: A utility class that simplifies sending progress updates with percentages and messages
3. **Wrapped Handlers**: Tool registration automatically injects the server context into handlers
### Example Usage in Tools
```typescript
import { ProgressReporter } from "../utils/progress.js";
import { ToolHandlerContext } from "../types/index.js";
handler: async (params: ParamType, context: ToolHandlerContext = {}) => {
const progress = new ProgressReporter(context);
await progress.report(25, "Fetching data from API...");
// Do work
await progress.report(75, "Analyzing results...");
// More work
await progress.report(100, "Complete!");
return results;
};
```
The `generate_seo_report` tool demonstrates this pattern with 6 major progress checkpoints that provide detailed status updates throughout the multi-step analysis process.
## Usage Analytics & Privacy
To improve the product, this MCP server records which tools are used (tool names, errors, and latency). This helps us understand which features are useful and where things break. We do **not** train any models on your data — see the [InfraNodus terms and conditions](https://infranodus.com/terms-conditions).
If you prefer these usage stats not to be linked to your InfraNodus account, set the `MCPCAT_ANONYMOUS` environment variable:
```json
{
"mcpServers": {
"infranodus": {
"command": "npx",
"args": ["-y", "infranodus-mcp-server"],
"env": {
"INFRANODUS_API_KEY": "YOUR_INFRANODUS_API_KEY",
"MCPCAT_ANONYMOUS": "1"
}
}
}
}
```
With `MCPCAT_ANONYMOUS=1` set, tool usage is still recorded but stays anonymous — it is not tied to your user account. It also disables the per-user `submit_workflow_feedback` log described under tool 37 (the report includes duration / error / retry of the call it rates).
### Project learnings
The `enable_project_learnings`, `add_project_learnings`, and `get_project_learnings` tools let the assistant keep an append-only graph of what it learned about operating in a project (`learn-<project>` in your InfraNodus account). They are opt-in per project: the graph exists only after you explicitly ask to enable it, the write tool refuses when the graph is missing and never creates one, and every batch is shown to you before it is saved. If you never want these tools exposed, set `INFRANODUS_LEARNINGS=0` in the server's environment (same place as `MCPCAT_ANONYMOUS`). To stop collecting for a project, delete its `learn-<project>` graph in InfraNodus.
As most tools have the `doNotSave=1` option turned on, all the graphs you create are ephemeral and won't be saved into your InfraNodus account unless you explicitly ask the MCP server to do so. We will also not keep the logs of the data you process. We also never use your data for LLM training according to the InfraNodus [https://infranodus.com/terms-conditions](terms of service).
## Troubleshooting
### Server doesn't appear in Claude
1. Verify the configuration file path is correct
2. Check that the API key is valid
3. Ensure Node.js is in your system PATH
4. Restart Claude Desktop completely
### API Key Issues
- Get your API key at: [https://infranodus.com/api-access](https://infranodus.com/api-access)
- Ensure the key is correctly set in the configuration
- Check that the key has not expired
### Build Errors
```bash
# Clean install
rm -rf node_modules package-lock.json
npm install
npm run build
```
## Resources
- [InfraNodus Website](https://infranodus.com)
- [InfraNodus MCP](https://infranodus.com/mcp)
- [InfraNodus API Documentation](https://infranodus.com/api-access)
- [MCP Protocol Documentation](https://modelcontextprotocol.io)
- [Graph Theory Concepts](https://noduslabs.com/research/)
## License
MIT
## Support
For issues related to:
- This MCP server: Open an issue in this repository
- InfraNodus API: Contact support@infranodus.com
- MCP Protocol: Visit the [MCP community](https://modelcontextprotocol.io)
This server cannot be deployed
Maintenance
ActivityActive
ResponsivenessLimited data