# Knowledge Graph Tools: Which One Fits Your Notes, Your Data or Your AI

> Knowledge graph tools sorted by job: personal note graphs, graph databases, visualization and analysis, AI graph builders and agent memory, with open-source and free options marked.

- Source: https://ivgraph.com/journal/knowledge-graph-tools/
- Author: [Andrii Danylchenko](https://ivgraph.com/team/#Andrii-Danylchenko), Founder / CTO, IVGraph
- Published: 2026-11-03
- Type: Guide
- About: Knowledge graph ([Wikidata Q33002955](https://www.wikidata.org/wiki/Q33002955))
- Mentions: Graph database ([Wikidata Q595971](https://www.wikidata.org/wiki/Q595971)); Graph drawing ([Wikidata Q739462](https://www.wikidata.org/wiki/Q739462)); [GraphRAG](https://microsoft.github.io/graphrag/); Model Context Protocol ([Wikidata Q133436854](https://www.wikidata.org/wiki/Q133436854)); Neo4j ([Wikidata Q1628290](https://www.wikidata.org/wiki/Q1628290)); Gephi ([Wikidata Q5548660](https://www.wikidata.org/wiki/Q5548660)); Obsidian ([Wikidata Q103994532](https://www.wikidata.org/wiki/Q103994532)); [IVGraph](https://ivgraph.com/)

## TL;DR

**There is no single best knowledge graph tool, only the right one for the job.** To see your notes as a graph, use Obsidian, Logseq or Anytype, or IVGraph if your notes live in Notion. To store a graph and query it, use a graph database such as Neo4j or GraphDB. To explore and measure a network, use Gephi or Cytoscape. To build a graph from documents with AI, try Neo4j’s LLM Knowledge Graph Builder, Microsoft GraphRAG or LightRAG. To give an AI agent memory, use the MCP memory server or Graphiti. Every job has a free or open-source option.

Knowledge graph software covers very different products: note apps that draw your links, databases that answer questions across millions of relationships, desktop tools for network analysis, AI pipelines that turn documents into graphs, and memory servers for AI agents. So the useful question is not which tool is best, but which job you need done. Below, 18 tools are sorted by job and checked on one date for license, free option, price and whether you need to write code. Then come five questions for choosing and three ways to try one on your own data.

## Five jobs a knowledge graph tool can do

A [knowledge graph](https://ivgraph.com/journal/what-is-a-knowledge-graph/) stores things as nodes and the named relationships between them as edges, so people and software can follow connections instead of searching through text. Our guide includes [a small graph you can click](https://ivgraph.com/journal/what-is-a-knowledge-graph/#try-it). Tools differ in which part of the work they take on:

1. **Map your notes.** See the links you already make between notes, pages and ideas: what is central, what is isolated. Obsidian, Logseq, Anytype, InfraNodus, Kumu, and IVGraph for Notion. See [tools for your notes](https://ivgraph.com/journal/knowledge-graph-tools/#notes).
2. **Store and query.** Load records into a database and ask questions that cross several links, such as “which customers bought products made with this supplier’s parts?” Neo4j, Memgraph, FalkorDB, GraphDB, Amazon Neptune. See [graph databases](https://ivgraph.com/journal/knowledge-graph-tools/#databases).
3. **Visualize and analyze.** Lay out a network you already have, find clusters and hubs, measure who sits in the middle. Gephi, Cytoscape, yEd, Neo4j Bloom. See [visualization tools](https://ivgraph.com/journal/knowledge-graph-tools/#visualize).
4. **Build from text with AI.** Let a language model extract entities and relationships from documents. Neo4j LLM Knowledge Graph Builder, Microsoft GraphRAG, LightRAG. See [AI graph builders](https://ivgraph.com/journal/knowledge-graph-tools/#ai).
5. **Remember for an agent.** Give an AI agent a graph where it saves facts and reads them back in later sessions. The MCP memory server, Graphiti, mem0. See [agent memory](https://ivgraph.com/journal/knowledge-graph-tools/#memory).

The jobs overlap: Neo4j stores graphs and ships a visual explorer, and LightRAG builds a graph from text and lets you browse it. Choose by the job you need most.

## Knowledge graph tools at a glance

> **Checked October 28, 2026.** Licenses come from each project’s repository or license page, prices from each vendor’s pricing page on that date, in the vendor’s currency and before tax. A dash means no paid plan is listed. Free tiers and prices change often, so check the tool’s own page, linked in its section below, before you decide. “Source available” means you can read and run the code, but the license is not approved by the Open Source Initiative and limits some uses.

| Tool | Best for | License | Free option | Paid from | Code needed |
| --- | --- | --- | --- | --- | --- |
| **Obsidian** | Linked notes in local Markdown files | Closed source | The full app, also for work | Sync $4/mo, billed yearly | No |
| **Logseq** | Outlines and daily journals | Open source (AGPL-3.0) | Everything | — | No |
| **Anytype** | Private notes as typed objects | Source available | Free plan | $4/mo (Plus) | No |
| **InfraNodus** | Topics and gaps in notes and texts | Closed source | 3 graphs | €12/mo, billed yearly | No |
| **Kumu** | Relationship maps you share | Closed source | Public projects | $9/mo per private project | No |
| **IVGraph** | A Notion workspace as a 3D graph | Closed source | Up to 500 nodes | $5/mo (Plus) | No |
| **Neo4j** | Apps and analytics on a property graph | Community Edition open source (GPLv3) | Community Edition, AuraDB Free | $0.09 per GB-hour (AuraDB Professional) | Yes, Cypher |
| **Memgraph** | Real-time and streaming data | Source available (BSL 1.1) | Community Edition | Quote (Enterprise) | Yes, Cypher |
| **FalkorDB** | A fast backend for GraphRAG and agents | Source available (SSPL) | Self-hosted, cloud free tier | $73/mo (1 GB cloud) | Yes, Cypher |
| **GraphDB** | RDF data, ontologies and reasoning | Closed source | GraphDB Free (1 CPU core) | Quote (Enterprise) | Yes, SPARQL |
| **Gephi, Gephi Lite** | Visual network analysis | Open source (GPL-3.0) | Everything | — | No |
| **Cytoscape** | Networks with many attributes | Open source (LGPL-2.1) | Everything | — | No |
| **yEd** | Clean diagrams with automatic layout | Closed source | Everything, also commercially | — | No |
| **LLM Knowledge Graph Builder** | Trying text-to-graph on your files | Open source (Apache 2.0) | Hosted app with AuraDB Free | — | No |
| **Microsoft GraphRAG** | Questions about a whole document set | Open source (MIT) | Everything; you pay for model calls | — | Yes, Python |
| **LightRAG** | GraphRAG with a web interface | Open source (MIT) | Everything; you pay for model calls | — | Setup, then a web UI |
| **Graphiti** | Agent memory that tracks change over time | Open source (Apache 2.0) | Everything; you run the database | — | Yes, Python |
| **MCP memory server** | A small local memory graph for Claude | Open source (MIT, moving to Apache 2.0) | Everything | — | One command |

The pattern: note tools cost little and need no code. Databases and AI pipelines are often free to download, but they expect queries or Python, and you pay for hosting and model calls.

### Open-source knowledge graph tools

These projects use licenses approved by the Open Source Initiative:

- **Notes:** Logseq (AGPL-3.0).
- **Databases:** Neo4j Community Edition (GPLv3) and [ArcadeDB](https://github.com/ArcadeData/arcadedb) (Apache 2.0), a multi-model database that answers SQL, Cypher and Gremlin queries.
- **Visualization:** Gephi and Gephi Lite (GPL-3.0), Cytoscape (LGPL-2.1), and for developers Cytoscape.js and sigma.js (MIT).
- **AI builders:** the Neo4j LLM Knowledge Graph Builder (Apache 2.0), Microsoft GraphRAG and LightRAG (MIT).
- **Agent memory:** Graphiti (Apache 2.0) and the MCP memory server, whose project is moving its code from MIT to Apache 2.0.

Memgraph and FalkorDB call themselves open source, but they use the Business Source License 1.1 and the Server Side Public License, and Anytype uses its own Any Source Available License. You can read and run the code; read the license before you build a commercial service on it. One more warning: [Kùzu](https://github.com/kuzudb/kuzu), an embedded graph database that older lists still recommend, was archived on October 10, 2025, and Graphiti now marks its Kùzu driver as deprecated. Don’t start new work on it.

## For your notes: personal knowledge graph tools

A personal knowledge graph is your own material, such as notes, people, projects, sources and ideas, connected by links you make. For most people, the right knowledge graph app is the note app they already use, as long as it can show those links as a graph. The apps differ mainly in what counts as a link.

### Obsidian graph view

[Obsidian](https://obsidian.md/pricing) keeps notes as Markdown files in a local folder, and its graph view is built in: every note is a node, every internal link a line. The global graph shows the whole vault. The local graph shows the notes around the one you are reading, with a depth slider. Groups color notes by search queries such as `path:Projects` or `tag:#book`. The app is free, including for work; Sync costs $4 a month per user, billed yearly. Our [Obsidian graph view guide](https://ivgraph.com/journal/obsidian-graph-view/) explains every setting. It suits people who write alone in local files.

### Logseq

[Logseq](https://logseq.com/) is a free, open-source outliner (AGPL-3.0) built around daily journals. Its graph draws links between pages, not blocks: if a journal entry mentions two pages, both connect to that day, not to each other. Logseq 2.0, which stores notes in a database instead of Markdown files, is in beta; [version 2.0.2](https://github.com/logseq/logseq/releases/tag/2.0.2) came out on October 7, 2026. For other apps in this group, see [Obsidian alternatives](https://ivgraph.com/journal/obsidian-alternatives/).

### Anytype

[Anytype](https://anytype.io/pricing) keeps your data on your device and treats everything as typed objects, such as notes, tasks and people. In its graph every object is a node, and lines come from links in the text, from properties that point to other objects and from backlinks; you can filter by object type. The code is published under the Any Source Available License. The free plan includes 100 MB of remote storage for sync; Plus costs $4 a month. Compared with Notion: [Anytype vs Notion](https://ivgraph.com/journal/anytype-vs-notion/).

### InfraNodus

[InfraNodus](https://infranodus.com/obsidian-plugin) builds a graph from the text itself rather than from your links: concepts that appear together in your text become connected. It then shows topical clusters, the concepts that hold them together and the gaps between clusters, and its AI suggests research questions that could bridge a gap. It works on texts, files and web sources, as an Obsidian plugin (InfraNodus AI Graph View) and through an MCP server for agents. The free plan includes 3 graphs; paid plans start at €12 a month billed yearly, with a 14-day trial. It helps you see what your notes are about and what they miss.

### Kumu

[Kumu](https://kumu.io/pricing) is a web app for relationship maps, drawn by hand or imported from a spreadsheet, then made readable with decorations, filters and clustering. Public projects are free; a private project costs $9 a month. It suits maps you share more than private notes.

### IVGraph for Notion

Notion has no built-in graph view, and that is the gap we built [IVGraph](https://ivgraph.com/) for. It is our product, so weigh this card with that in mind. You sign in with Notion and give read-only access, and IVGraph draws the workspace as an interactive 3D graph: pages and databases are nodes, while mentions, backlinks, database relations and parent pages are links. You can switch between 3D, Sphere, Radial and Flat views, search pages with results grouped by database, rank pages by importance with PageRank, spot the pages nothing links to, and color, resize or hide pages with style rules. On the Pro and Max plans, a read-only [API for AI agents](https://ivgraph.com/docs/v1-llm-api/) lets Claude or your own script search the workspace, read pages and follow links.

![IVGraph’s Sphere view: every page and database of a sample Notion workspace placed on a globe, with the layout menu on the left](https://ivgraph.com/assets/img/shots-2026-10/v3-dense-c-1440w.webp)

*A sample Notion workspace in IVGraph’s Sphere view: 3,389 nodes and 6,107 links. The menu on the left switches between the 3D, Sphere, Radial and Flat layouts.*

The app works only with Notion; for an Obsidian vault, our separate free [Obsidian Graph Online](https://ivgraph.com/tools/obsidian-graph/) tool draws the graph in your browser. IVGraph does not import CSV or RDF, and it is not a database or a query language. The Free plan covers up to 500 nodes; Plus is $5 a month for 3,000 nodes, Pro $10 for 10,000 nodes and the API, and you can try Pro for a week for $1.99. For what Notion offers on its own, see [does Notion have a graph view](https://ivgraph.com/journal/does-notion-have-graph-view/).

## Graph databases: store and query a knowledge graph

A knowledge graph database earns its place when you need to ask questions, not only look: what depends on this service, or which customers are linked to a flagged account through shared devices. You model the data as nodes and relationships, load it, and query it in a language such as Cypher or SPARQL. Every option here expects someone comfortable with queries or code.

### Neo4j

[Neo4j](https://neo4j.com/pricing/) stores a property graph and is queried with Cypher, which Neo4j describes as GQL-compliant. The Community Edition is open source under GPLv3 and free to run on your own server. In the cloud, AuraDB Free needs no payment method and limits the number of nodes and relationships; AuraDB Professional starts at $0.09 per GB-hour with a 1 GB minimum, roughly $66 for a month of a 1 GB instance. [Bloom](https://ivgraph.com/journal/knowledge-graph-tools/#visualize) and the [LLM Knowledge Graph Builder](https://ivgraph.com/journal/knowledge-graph-tools/#ai), covered below, come from Neo4j too.

### Memgraph

[Memgraph](https://memgraph.com/pricing) keeps the graph in memory, writes it to disk for durability and also speaks Cypher. The free Community Edition already includes replication, vector search, stream connectors for Kafka, Pulsar and Redpanda, and the MAGE library of graph algorithms. The code is under the Business Source License 1.1, and Memgraph Lab is its visual interface. Enterprise pricing is a quote based on memory size. It fits data that arrives in streams and must be queried right away.

### FalkorDB

[FalkorDB](https://www.falkordb.com/pricing/) is built on Redis and stores graphs as sparse matrices, using the GraphBLAS library; you query it with openCypher. It targets GraphRAG and agent memory: it has its own GraphRAG SDK and is one of the databases Graphiti supports. One Docker command starts the database with a browser interface on your machine. The code is under the Server Side Public License. The cloud has a free tier, and paid plans start at $73 a month for 1 GB.

### GraphDB

[GraphDB](https://graphwise.ai/components/graphdb/) is an RDF database: data is stored as subject–predicate–object triples, queried with SPARQL, and the database can infer new facts from an ontology using the RDFS and OWL 2 RL and QL rulesets. Ontotext built it; in 2024 Ontotext and Semantic Web Company merged into Graphwise, which sells it now. GraphDB Free runs on one CPU core and, since version 11, needs a free license key sent by email. The Enterprise Edition is priced on request.

### Amazon Neptune and Stardog

[Amazon Neptune](https://aws.amazon.com/neptune/pricing/) is a managed graph database on AWS that supports Gremlin, openCypher and SPARQL, so it can hold a property graph or RDF data. New users get a 30-day free trial on a small instance; after that you pay per use. It suits teams already on AWS. [Stardog](https://www.stardog.com/pricing/), an RDF-based enterprise platform, offers Stardog Free: allowed for commercial use but not open source, renewed yearly, without high availability or backups.

### Property graph or RDF?

A property graph (Neo4j, Memgraph, FalkorDB) stores nodes and relationships with properties on both. You design the model around your own questions, which suits applications and GraphRAG backends. RDF (GraphDB, Stardog) stores triples with global identifiers, so data from different sources and shared vocabularies merge more easily, and an ontology lets the database infer facts you did not write down. Building an app or an AI backend? Start with a property graph. Connecting data across teams or organizations, or publishing linked data? Look at RDF. Our guide explains [triples, schemas and ontologies](https://ivgraph.com/journal/what-is-a-knowledge-graph/#building-blocks).

## Knowledge graph visualization: explore and analyze a graph

Visualization tools take a graph you already have, as a file, a spreadsheet of links or a database, and help you see its structure: clusters, hubs and the nodes that bridge groups. Some also measure it. A tidy picture of bad data is still bad data.

### Gephi and Gephi Lite

[Gephi](https://gephi.org/) is a free, open-source (GPL) desktop app for visual network analysis. It has algorithms for community detection, centrality and clustering, layouts such as ForceAtlas2, a spreadsheet view for editing data, and export to PNG, PDF and SVG. Version 0.11, released in May 2026, brought a major performance upgrade; 0.11.3 came out on September 6, 2026. [Gephi Lite](https://lite.gephi.org/) is the lighter web version: it runs in a browser with nothing to install and is built on the sigma.js library.

### Cytoscape

[Cytoscape](https://cytoscape.org/) is an open-source (LGPL) Java desktop app for network analysis. It began in 2003 as a tool for biomolecular interaction networks; today its apps cover bioinformatics, social network analysis and the semantic web. Version 3.10.5 came out on September 29, 2026. Its sibling [Cytoscape.js](https://js.cytoscape.org/) (MIT) is a library for putting graphs into web apps.

### yEd

[yEd](https://www.yworks.com/products/yed) is a free desktop diagram editor from yWorks, free for commercial use as well but not open source. Its strength is automatic layout: import nodes and edges from an Excel sheet or XML, and its layout algorithms arrange even large diagrams. Version 3.25.1 is current, and yEd Live runs in a browser. Use it for a clean diagram, not for analysis.

### Neo4j Bloom

If your graph lives in Neo4j, [Bloom](https://neo4j.com/docs/aura/explore/introduction/) lets people who don’t write Cypher explore it: search with near-natural-language phrases, expand neighbors, inspect and edit records. It is part of the Aura console, where Neo4j lists access to all graph tools on AuraDB Free, and also exists as a standalone app.

### Commercial tools for large graphs and investigations

- [Graphistry](https://www.graphistry.com/get-started) renders very large graphs on GPUs and works from Python notebooks through PyGraphistry. Graphistry Hub has a free tier; Professional costs $83 a month billed yearly, or $100 month to month.
- [KeyLines](https://cambridge-intelligence.com/keylines/) and ReGraph from Cambridge Intelligence are commercial JavaScript and React SDKs for building graph visualization into your own product; trials are on request.
- [Linkurious](https://linkurious.com/pricing/) Enterprise runs on top of a graph database for investigation teams, with search, a no-code query builder and collaboration. The cloud edition costs €490 per user per month after a 30-day trial.

### For developers: graph visualization libraries

To put a graph inside your own app, use a library instead of an application:

- **Cytoscape.js** (MIT): graph theory and visualization in one library, with many layout extensions.
- **[sigma.js](https://github.com/jacomyal/sigma.js)** (MIT): WebGL rendering for graphs with thousands of nodes and edges, now at version 4.
- **[D3](https://github.com/d3/d3)** (ISC): low-level building blocks, including a force simulation, for fully custom drawings.
- **[3d-force-graph](https://github.com/vasturiano/3d-force-graph)** (MIT): a 3D force-directed graph on Three.js and WebGL. IVGraph’s 3D view is built on it.

## Build a knowledge graph from text with AI

A knowledge graph builder reads documents and asks a language model to extract entities, such as people, products and concepts, and the relationships between them. The result is stored as a graph you can query or give to a chatbot. Treat the output as a draft: models merge different things that share a name, miss relationships and name the same relationship in different ways. Check a sample by hand before you rely on it. For when graph retrieval beats ordinary vector search, see our [GraphRAG explainer](https://ivgraph.com/journal/graph-rag/).

### Neo4j LLM Knowledge Graph Builder

The [LLM Knowledge Graph Builder](https://neo4j.com/labs/genai-ecosystem/llm-graph-builder/) is a quick way to try the idea on your own files. It is an online app from Neo4j Labs: connect a Neo4j database, add PDFs, documents, web pages or YouTube transcripts, choose a model, and it builds two layers. One holds your documents split into chunks with embeddings; the other holds entities and their relationships, each linked to the chunk it came from. Then you can chat with the result and open it in Bloom. Neo4j notes that it works best on long English text, less well on spreadsheets, images and slides, and better when you define the node and relationship types. The code is under Apache 2.0, version 0.8.7 came out in July 2026, and Docker Compose runs it locally.

### Microsoft GraphRAG

[Microsoft GraphRAG](https://github.com/microsoft/graphrag) (MIT; version 3.3.0 from October 8, 2026) is a Python pipeline that extracts entities and relationships, groups them into communities and writes summaries, so it can answer questions about a whole collection as well as about specific entities. Its README warns that indexing can be expensive and recommends tuning the prompts for your data. It also calls the code a demonstration, not an officially supported Microsoft product. Start with a small sample and watch the model bill.

### LightRAG

[LightRAG](https://github.com/HKUDS/LightRAG) (MIT; version 1.5.7 from September 2, 2026) is a research project, presented at EMNLP 2025, that builds a graph index from your documents for retrieval. It comes with a server and a web interface for adding documents, asking questions and browsing the graph, and it can store data in Neo4j, PostgreSQL, MongoDB or OpenSearch. Choose it when you want GraphRAG with an interface out of the box.

**Graphify is a code graph, not a Notion tool.** If you searched for a Graphify alternative expecting a Notion graph: in 2026, search results for that name are about [Graphify](https://github.com/Graphify-Labs/graphify), an open-source (Apache 2.0) skill for Claude Code, Cursor, Codex and Gemini CLI. It turns a codebase, with its docs, SQL schemas and PDFs, into a knowledge graph the coding agent can query.

## Graph memory for AI agents

An AI agent forgets everything between sessions unless something stores what it learned. Graph memory stores it as entities and relationships, so the agent can look up a person and follow links to their projects and decisions instead of rereading old chats. Three options cover most setups. Our [Claude Code memory guide](https://ivgraph.com/journal/claude-code-memory/) shows where each fits next to CLAUDE.md.

### The MCP memory server

The Model Context Protocol project’s [memory server](https://github.com/modelcontextprotocol/servers/tree/main/src/memory) (`@modelcontextprotocol/server-memory`, version 2.0.0 from October 19, 2026) keeps a small knowledge graph in a local file: entities with a name, a type and observations, and directed relations named in the active voice, such as `works_at`. Claude or any other MCP client reads and writes it through nine tools, from `create_entities` to `search_nodes`. Search matches words, not meaning. It is free, runs on your computer and fits one person’s assistant better than a team.

### Graphiti

[Graphiti](https://github.com/getzep/graphiti) (Apache 2.0; version 0.31.2 from October 20, 2026) builds a temporal graph: every fact has a validity window, so when something changes the old fact is marked invalid instead of deleted, and you can ask what was true at any point. Each fact links back to the message, document or record it came from. It needs a graph database (Neo4j, FalkorDB or Amazon Neptune) and a model that supports structured output; OpenAI is the default. The repository includes a reference MCP server, and its README warns that it has no authentication, so keep it off untrusted networks. Zep, the company behind Graphiti, sells a managed enterprise service built on it.

### mem0

[mem0](https://github.com/mem0ai/mem0) (Apache 2.0) is a memory layer for agents: it extracts short memories from conversations and retrieves the relevant ones later. Its graph memory, which links memories through the people, places and concepts they mention, runs on the hosted Mem0 Platform and is listed under the Custom plan on its [pricing page](https://mem0.ai/pricing). The free plan gives 10,000 additions and 1,000 retrievals a month for the first 30 days, then 50,000 and 10,000 with your own model key.

If the knowledge your agent needs already lives in Notion, you may not need a separate memory at all. IVGraph’s read-only API on Pro and Max lets an agent search a workspace, read a page as Markdown and follow its links.

## How to choose in five questions

Each answer points to a section above.

1. **What is your data?** Notes you write: a note app with a graph ([notes](https://ivgraph.com/journal/knowledge-graph-tools/#notes)). Documents you collected: an AI builder ([AI](https://ivgraph.com/journal/knowledge-graph-tools/#ai)). Records from spreadsheets or systems: a graph database ([databases](https://ivgraph.com/journal/knowledge-graph-tools/#databases)).
2. **Do you want to look or to ask?** Looking for clusters, hubs and orphans needs a note app or a visualizer. Asking “who is connected to whom, and through what?” needs a database and a query language.
3. **Just you or a team?** Alone: Obsidian, Logseq, Anytype or Gephi. A team: Kumu for shared maps, or a hosted database such as AuraDB or Neptune with Bloom or Linkurious on top.
4. **Cloud or your own machine?** On your machine: Obsidian, Logseq, Anytype, Gephi, Cytoscape, yEd, Neo4j Community Edition, FalkorDB in Docker and the MCP memory server. In the cloud: InfraNodus, Kumu, IVGraph, AuraDB, Neptune and Mem0 Platform. Any tool that sends text to a model provider sends your data there too.
5. **Will you write code?** No: the note apps, InfraNodus, Kumu, IVGraph, Gephi, yEd and the LLM Knowledge Graph Builder. A little: the MCP memory server and LightRAG. Yes: the databases, GraphRAG and Graphiti.

Usually two or three tools survive these questions. Try them on a small, real sample, such as one project or fifty documents, and keep the one that answers your actual questions faster.

## Try a knowledge graph on your own data in 10 minutes

Three short paths, one for each kind of data. All three start free.

### Notes in Obsidian

1. Open your vault and click **Open graph view** in the ribbon, or run **Graph view: Open graph view** from the command palette.
2. Open a note you know well, run **Open local graph** and set the depth to 2.
3. In the graph settings, turn on **Orphans** and look at the notes that float alone. Each one is a link you never made.

### A Notion workspace in IVGraph

1. Click **Login with Notion** on ivgraph.com and approve access in Notion’s own window.
2. Wait for the first sync. A small workspace takes about two minutes.
3. Open the graph, search for a project to see everything linked to it, then try the Sphere layout.

![A search for “sync” in IVGraph: 50 matching Notion pages grouped by database next to the 3D graph of the workspace](https://ivgraph.com/assets/img/shots-2026-10/graph-search-1440w.webp)

*Search in IVGraph groups matching pages by database, so a project, its meeting notes and its tasks show up side by side next to the graph.*

### Documents in Neo4j

1. Create an **AuraDB Free** instance at [console.neo4j.io](https://console.neo4j.io/) and download its credentials file. No payment method is needed.
2. Open the [LLM Knowledge Graph Builder](https://llm-graph-builder.neo4jlabs.com/) and drop the credentials file into its connection dialog.
3. Add two or three PDFs or a web page link, choose a model and generate the graph. Then ask a question in the chat and check which chunks the answer used.

Whichever path you take, judge the result by one question you really have. If the graph answers it faster than search does, keep going. If it doesn’t, the problem is usually missing links in your data, not the tool.

## FAQ

### What is the best knowledge graph software?

There is no single best one; it depends on the job. For your own notes, use Obsidian, Logseq or Anytype, or IVGraph if your notes are in Notion. To store and query a graph, use a graph database such as Neo4j, which has a free cloud tier and a free Community Edition, or GraphDB for RDF data. For visual analysis, use Gephi or Cytoscape. To build a graph from documents with AI, try Neo4j’s LLM Knowledge Graph Builder, Microsoft GraphRAG or LightRAG. For agent memory, use the MCP memory server or Graphiti.

### Are there free or open-source knowledge graph tools?

Yes. Open source: Logseq, Neo4j Community Edition, ArcadeDB, Gephi and Gephi Lite, Cytoscape, the Neo4j LLM Knowledge Graph Builder, Microsoft GraphRAG, LightRAG, Graphiti and the MCP memory server. Free but not open source: Obsidian, yEd, GraphDB Free, Stardog Free, Neo4j AuraDB Free and IVGraph up to 500 nodes. Memgraph, FalkorDB and Anytype publish their code under source-available licenses that restrict some commercial uses.

### What tool can I use to build a personal knowledge graph?

Start with a note app that links notes and draws those links as a graph: Obsidian, Logseq or Anytype. If your notes are in Notion, IVGraph shows the workspace as a 3D graph of pages, mentions and database relations. To analyze the text itself for topics and gaps, add InfraNodus. Personal notes rarely need a graph database.

### What is the difference between a knowledge graph tool and a graph database?

A graph database is one kind of knowledge graph tool: it stores nodes and relationships and answers queries written in languages such as Cypher or SPARQL. Knowledge graph tools also include note apps that draw your links, visualizers such as Gephi, AI builders that extract a graph from text, and memory servers for AI agents. Many projects combine several of them.