InfraNodus

Turn text into a knowledge graph with topic clustering and graph-grounded AI insight through InfraNodus. Agents find the gaps and blind spots in a body of feedback or research rather than just summarizing it.

14 actions API key available

InfraNodus website ↗ Platform Documentation ↗ Capability data verified 2026-08-01

A quarter of customer interview transcripts and support tickets closes out

Generate Graph And Statements builds the knowledge graph and returns clusters, ranked concepts and structural gaps

Generate AI Advice For Text returns bridging questions aimed at the gaps rather than a summary

Compare Search Results And Intent tests which of those gaps also show up as unanswered search demand

Agent posts the ranked gaps with their bridging concepts and the demand signal beside each

The content owner marks which gaps are real briefs and which are absent for a good reason

What This Integration Enables

Most text analysis tools tell you what a body of writing says. InfraNodus tells you what it does not say, which is a genuinely different question and a much harder one to answer by reading. It builds a knowledge graph from a text, a set of statements or a saved graph, and returns it in Graphology JSON alongside network analytics: topical clusters, top-ranked concepts, structural content gaps where clusters fail to connect, and conceptual gateways that bridge them. The AI advice actions are grounded in that graph structure rather than in the raw text, so they surface research questions and blind spots an ordinary summariser will not, steered by an optimisation strategy you choose: develop the discourse, reinforce it, bridge structural gaps, or explore latent and peripheral topics.

There is a second, quieter capability that matters to anyone building retrieval flows. Generate DOT Graph From Text and Generate DOT Graph From Graph condense a graph into Graphviz DOT plus a human-readable summary, carrying the same relational information as the JSON in a fraction of the tokens. Injecting that block into a prompt gives a model the structure of a corpus without paying for the corpus, which is a materially better use of a context window than more chunks.

The search actions apply the same idea to content strategy. Build Search Results Graph maps informational supply by graphing the top Google results. Build Search Intent Graph maps demand by graphing the keywords people actually search around a query. Compare Search Results And Intent returns the difference: the relations that exist in intent and not in published results, which is a direct read on what people look for and fail to find. Nothing is persisted to your InfraNodus account by default, and saving a graph is an explicit choice.

Without FlowRunner

Feedback reduced to a summary A model reads the corpus and returns the themes that were already obvious

Topic coverage judged by memory Whether something has been written about is answered by whoever has been here longest

Keyword research is a flat list A spreadsheet of terms with volumes, and no picture of how they relate

With FlowRunner

Feedback rendered as structure Topical clusters, ranked concepts and the gaps between clusters come back as a graph

Coverage measured against the corpus Absence is detected structurally, from what the graph does not connect

Demand and supply compared directly The difference between what people search for and what exists is a graph, not two lists

Use Case Scenarios

Finding the blind spot in a quarter of customer feedback

Support tickets, interview transcripts and churn survey responses accumulate faster than anyone reads them, and a summary of them tends to return the themes the team already discusses. The agent runs Generate Graph And Statements over the whole corpus and gets the topical clusters plus the structural gaps between them, then calls Generate AI Advice For Text with the bridge-gaps strategy to get up to three questions that connect clusters currently sitting apart. Those questions are the interesting output, because they name relationships nobody in the corpus articulated. The graph summary and the questions go into a Notion page for the quarterly review.

Structure instead of chunks in a retrieval prompt

A retrieval flow pulls a dozen documents for a question and would normally paste them into the prompt. Instead the agent calls Generate DOT Graph From Text on the retrieved set and passes the compact DOT block plus the graph summary alongside a much smaller selection of source passages. The model receives the shape of the relationships across all twelve documents and the exact wording of the few that matter, rather than a truncated pile of chunks where the relationships between documents were never represented at all.

Comparing two bodies of writing

Two things need comparing: what a competitor publishes and what you publish. The agent calls Generate AI Advice For Multiple Graphs with both corpora and the difference framing, and gets advice grounded in what one covers that the other does not. Run the same action against a set of analyst reports and a set of your own positioning documents and it answers a different question with the same mechanism, which is whether the way you describe your market matches the way the market describes itself.

Human-in-Loop Highlight

InfraNodus never writes anything to a customer, so the consequence does not live in the connector. It lives one step later, in what a team commits to on the strength of a gap. Compare Search Results And Intent will happily return twenty relations that exist in search intent and not in published results, and Generate Search Gap Advice will turn each into a plausible content angle. A structural gap means nobody has written about it, and that has two very different causes. Sometimes it is an underserved audience need and a real brief. Sometimes it is absent because the topic is wrong for the business, or already covered under vocabulary the graph did not connect, or a question people ask that nobody should answer. Graph structure cannot tell those apart, and neither can the agent.

So the agent does the analysis and stops before the commitment. It posts to the content owner: "Compare Search Results And Intent for 'invoice exception handling' returned 14 gap relations. Top 5 by intent weight listed with their bridging concepts and the clusters they would connect. Which of these become briefs?" The owner marks three, kills two, and adds context to one. The agent then runs Generate Search Gap Advice on only the approved three and drafts the briefs. The expensive judgment stayed with the person who has it, and the analysis that made the judgment cheap ran without them.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

14 actions

Knowledge Graphs

Generate Graph And Statements Builds a knowledge graph from a text, an array of statements or an existing saved InfraNodus graph, returning it in Graphology JSON along with network analytics: topical clusters, top-ranked concepts, structural content gaps and conceptual gateways. Optionally returns a human-readable graph summary suited to augmenting a retrieval prompt. Nothing is persisted to your account by default.
Generate DOT Graph From Text Converts plain text into a compact Graphviz DOT representation of its knowledge graph plus a human-readable summary. The DOT form carries the same relational information as the JSON graph in a fraction of the tokens, which makes it the preferred way to inject structural context into a model prompt.
Generate DOT Graph From Graph Condenses an existing Graphology JSON graph into a compact DOT representation plus a summary, optionally narrowed to specific concepts or topical clusters. Used to shrink a large graph into a prompt-sized block of structure without losing the relations that matter.
List Graphs Retrieves the knowledge graphs saved in your InfraNodus account, with filters for name and description text, context type, language, favourite status and creation date range. Comma-separated values within the search and type filters combine with OR, while different filter categories combine with AND.
Search Statements Searches the statements stored across your InfraNodus graphs for a term and builds a knowledge graph from every match, so you can see how a concept is discussed across the whole corpus rather than inside one document. The search can be scoped to named graphs and the resulting graph capped to a maximum node count.

AI Insights

Generate AI Advice For Text Analyzes a text as a knowledge graph and returns advice grounded in the graph's structure rather than in the raw text, which is how it surfaces research questions, blind spots and bridging ideas a summariser would miss. Choose an optimisation strategy to develop the discourse, reinforce it, bridge structural gaps, or explore latent and peripheral topics. Returns up to three suggestions along with the underlying graph.
Generate AI Advice From Graph Sends an existing Graphology JSON graph together with a prompt and returns advice grounded in that graph's structure and semantics. Used to query a graph built earlier without reprocessing the source text, to run a chat over the graph with prior messages as context, or to answer in a specific language.
Generate AI Advice For Multiple Graphs Analyzes several sources at once, mixing plain texts, saved graphs and statement arrays, and returns advice grounded in their combined structure or in the difference between them. Used to compare two bodies of writing and surface what one covers that the other does not, or to derive one set of research questions spanning several documents.

Search Analysis

Build Search Results Graph Runs a Google search for a query and builds a knowledge graph from the top results, revealing which topics the existing content supply already covers and where the structural gaps are. This is the supply side of the picture; pair it with Build Search Intent Graph to compare supply against demand.
Generate Search Results Advice Builds a graph of the top Google results for a query and returns advice grounded in it, such as research questions or content angles targeting the structural gaps in what is already published. Used to find an underserved angle before writing.
Build Search Intent Graph Builds a knowledge graph from the keywords people actually search around a query, sourced from Google's suggested searches and AdWords recommendations. This maps informational demand rather than supply, exposing the topical clusters and related keywords worth targeting.
Generate Search Intent Advice Builds a graph of the keywords people search around a query and returns advice grounded in it, surfacing content ideas that match current demand and the gaps between what people ask about.
Compare Search Results And Intent Builds a graph of the relations that exist in search intent but not in published search results for a query, exposing what people look for and fail to find. Returns the difference graph with its topical clusters and content gaps, which is the core signal for locating underserved niches.
Generate Search Gap Advice Compares search intent against published results for a query and returns advice grounded in the difference, producing content ideas and research questions aimed at what audiences look for but cannot currently find.

Frequently Asked Questions

What can FlowRunner do with InfraNodus?

FlowRunner agents can run Generate Graph And Statements, Generate AI Advice For Text, and Generate AI Advice From Graph in InfraNodus, plus 11 more actions.

Does connecting InfraNodus to FlowRunner require OAuth?

No. InfraNodus connects to FlowRunner with an API key, no OAuth flow required.

Can InfraNodus trigger a FlowRunner workflow automatically?

InfraNodus doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with InfraNodus

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us