Both turn your material into nodes and connections, and both work inside the AI tools you already use. Where they part is what the graph is for once it exists. Graphify’s is a reading of your files, written out fresh each time they change, while Filamental’s is a place you build, arrange and write in, which is still there as you left it next week.
Graphify is an open-source tool that reads a repository, along with its docs, schemas, configs and PDFs, and turns it into a knowledge graph you can query. Code is parsed directly, across dozens of languages, without a model and without anything leaving your machine, while documents, PDFs and images go through an LLM. What comes out is an interactive graph page, a JSON file and a written report on what it found.
It lives where developers already work. It installs as a skill in Claude Code, Cursor, Codex, Gemini CLI and a long list of others, it can run as an MCP server, and a git hook rebuilds the graph on every commit, so the map doesn’t fall behind the code. Ask it for the path between two things and it’ll trace it, and it exports to Neo4j and GraphML if you want to take the graph somewhere else.
For “what touches what in this repository, and where did that come from”, it’s a thoroughly sensible answer, and we’d point you at it.
Graphify’s graph is an output. It’s produced from the source, and when the source changes it’s produced again, which is exactly right for a codebase, where the code is the truth and the map should follow it.
A plan, an argument, an investigation or a client brief has no source to be read back from, because the structure is the work. Somebody has to decide what matters, which things belong together and what each connection means, and that’s what Filamental is for. You position things and the layout is kept, every node carries its own notes, and the connectors are kinds of relationship you named yourself, such as “funds”, “contradicts” or “reports to”, rather than one generic link.
It’s also a document. The same project opens in the Publisher as pages you can format, share as a link or present on a screen, so the thing you thought in is the thing you hand over, with no second version to keep in step.
A generated graph can tell you what’s in the folder, though it tends to be quieter about what you were planning to do with it.
You don’t have to choose, and plenty of people shouldn’t. Let Graphify keep the map of the code, and use Filamental for everything built on top of it: the roadmap, the architecture decision and the reasons behind it, the incident review, and the explainer for people who’ll never open the repository.
The two meet in the same places. Filamental’s MCP server works in Claude Code, Cursor and VS Code, so the agent that just asked Graphify how two modules connect can turn round and build the decision into your Filamental project in the same session. Ask it how two nodes in that project are connected and it reads back the shortest route, the same one Pathfinder lights up on your screen.
And if a colleague’s AI can’t reach your machine at all, Export for AI saves the whole project as one Markdown file that opens with a summary of how it’s built, which any chat will read.
Filamental has its own version of the first step. Point Folder Exploration at a folder of documents (Markdown, text, Word and PDF) and the AI you’ve connected reads them and builds a project. You approve the classifications and connectors before anything is read and see an estimate of the cost before it starts, and what you get at the end is a first draft that’s yours to rearrange.
It runs on your own AI account or a model on your own computer, so the reading is billed to you and the documents never pass through us. It doesn’t parse source code, and for a repository, Graphify is where we’d start. How Folder Exploration works.
Building, Folder Exploration, the MCP connection, Pathfinder and Export for AI are all free on the Personal plan, permanently, with no account and no card. The paid tier is $120 a year, or $12 a month, and it buys the routes out: more than one live link at a time, updating a link you’ve already sent, downloading the HTML file, and presenting.
Not really, because they do different jobs. Graphify reads a codebase and its documents and generates a knowledge graph you can query, rebuilding it when the code changes. Filamental is a 3D and 2D workspace where you build and arrange a graph yourself, write notes on every node and turn the result into a document you can share or present. Plenty of people will reasonably use both, Graphify to map the code and Filamental for the thinking and planning built on top of it.
No. Filamental doesn't parse source code. Its Folder Exploration reads folders of documents (Markdown, text, Word and PDF files) with the AI you've connected and builds a project from them. For mapping a repository, Graphify, which parses code directly across many languages, is where we'd start.
Yes, and that's the point of it. Every node is a Markdown file in a folder you choose, with its relationships in the front matter, and you can move, rename, reclassify, connect and write about anything. Positions are saved, so the layout you arrange is the layout you come back to, and nothing regenerates over your changes.
Only if you ask it to, and you can build a project entirely by hand. Folder Exploration, the built-in assistant and MCP connections all run on an AI account or a local model that you connect yourself, billed to you. Your documents go straight from your computer to that provider and never pass through us.
They overlap a good deal. Graphify installs as a skill in agents including Claude Code, Cursor, Codex and Gemini CLI, and can run as an MCP server. Filamental connects over MCP to Claude Desktop, Claude Code, Cursor, VS Code and LM Studio, has a built-in assistant that works with the major AI providers, and can export a whole project as one Markdown file for any chat.
Ask the agent you already use to build it into a Filamental project, then arrange it until it says what you meant.