Graphify
Draw an arrow from an agent to the Graphify card. It asks a map of your project how things connect: who calls this, how checkout reaches the database, what breaks if I change that. You watch each question travel across the map as it happens.
Built on graphify, the open-source code knowledge graph by Graphify Labs:github.com/Graphify-Labs/graphify
9
graph tools an agent gets with one arrow: questions, neighbours, paths, impact and more
≈12
seconds to map this website’s own code the first time (472 files) in our test; after an edit, only changed files are read again
0
bytes of your code sent anywhere. graphify parses locally with tree-sitter, no LLM, no telemetry
Every question lights up the map
Three agents ask the same Graphify card. The question appears, the files where the search starts ripple, and the traversal spreads hop by hop along the calls and imports it follows. Click a question to replay it, or a file to see what it is linked to.
Result
6 files visited · depth 1
- 0blogHref()lib/blog.ts:186
- 1PostPage()app/blog/[slug]/page.tsx:109
- 1Footer()components/Footer.tsx:138
- 1resourceGroups()components/Nav.tsx:162
- 1blogFeed()lib/blogFeed.ts:41
An illustrative project. The tools, the traversal and the shape of the answers are the app’s.
Two graphs, two jobs
NeuroSquad has two code-graph plugins. They answer different questions, and they work side by side on the same workspace.
Graphify: the map
- Ask in plain words, get the relevant part of the graph back in one call
- Subsystems found by clustering, so you see what belongs together
- Docs, Markdown sections and NOTE / WHY comments live in the same graph as the code
- Blast radius: everything a change to a symbol would touch
- A live visual map on the card, with every agent query drawn on it
Code Graph: the precision tool
- Exact symbol search, callers and callees several hops deep
- Source of one function without opening the file
- Read-only Cypher for multi-step questions
- Semantic code search and uncommitted changes mapped to symbols
- The engine runs next to the code, also in WSL and on SSH hosts
Use Graphify to understand a codebase and plan a change, Code Graph to navigate to the exact lines. Arrow an agent to both: each tool’s description says which questions it is for.
About Code Graphgraphify’s own queries, scoped to your workspace
The tools of graphify’s MCP server, answered by graphify’s own code on this workspace’s graph. Each description says when the graph beats a text search.
Ask
graphify_query turns a question or a few symbol names into the relevant subgraph: files with line numbers, calls, imports, inheritance, each edge marked EXTRACTED or INFERRED.
Neighbours
graphify_neighbors lists everything that calls, imports or references one symbol, with the file and line of each use. graphify_node says where it lives and which subsystem it belongs to.
Paths and impact
graphify_path finds the shortest chain between two things. graphify_affected walks back from a symbol to everything that a change to it would touch.
The big picture
graphify_community lists one subsystem, graphify_god_nodes the most connected abstractions, graphify_stats the size of the graph.
Agents learn when the graph beats grep
A tool an agent never calls does nothing. The card tells connected agents what the graph is for, in plain facts, and never forces it.
A note on the first turn
The first turn with the arrow carries a short note: what is in the graph, which tool answers which question, and when text search is still the right tool.
Reminders where they help
Later turns get a one-line reminder when the prompt asks about code structure, and every few turns otherwise. Removing the arrow sends one note that the tools are gone.
A hint before a search
In Claude Code, a Grep, Glob or shell grep for what looks like a symbol gets one line pointing at graphify_neighbors. Quiet right after the agent used the graph, rate-limited otherwise.
Never a block
The hint is context, not a permission decision. Your allow and deny rules stay as they are and the search still runs. A switch on the card turns all of it off.
Turns carry the note in
Mapped as soon as you add the card
The card builds the full graph of the workspace folder right away and keeps it in step with your edits.
Progress you can see
Files read out of the total, the current step and the time left, on the card while the first build runs.
Agents are told the truth
While the first graph is being built, the tools answer with how far it is, so the agent uses its usual search meanwhile instead of failing.
Incremental updates
The card watches the folder and re-reads changed files a few seconds later; graphify’s cache keeps an update to seconds.
Your choice of files
.gitignore and .graphifyignore are respected. On the card you can leave out languages, add ignore patterns or include git-ignored generated code.
One card, one arrow
Everything happens on the canvas you already have open.
Add the card
Add-card menu → “Plugin…” → Graphify → “Add to canvas”.
Install once
Press Install. The app sets up graphify in its own folder (about 250 MB) and maps the whole project right away.
Draw an arrow
From an agent to the card. It gets the graph tools on its next tool listing, no restart for most CLIs.
Watch the questions
Each query plays on the map with its results. Click any file to open it in your editor.
Every CLI NeuroSquad runs
The tools come through NeuroSquad’s own MCP server, so every agent with an arrow can use them.
The graph tools: every agent NeuroSquad connects to its MCP server
Workspaces on this computer and in WSL (graphify reads the distro’s files from Windows). SSH workspaces are not supported; use Code Graph there.
Your code stays on your computer
graphify is a program on your machine, not a service.
Local parsing only
Code is parsed with tree-sitter on this computer. The card runs graphify’s code-only mode, with no LLM backend and no API key in its environment.
Nothing phoned home
graphify has no telemetry, and its optional query log is switched off by the app.
Pinned install
A private Python is installed with uv into the app’s folder. uv’s download and every Python package are pinned by checksum. Your own Python and settings are not used or touched.
Kept in the app’s folder
The graph lives in NeuroSquad’s data folder, nothing is written into your project, and deleting the card deletes its graph.
Open source inside
Powered by graphify
The extraction, the graph and the queries are graphify’s: tree-sitter parsing of about 40 languages, cross-file call and import resolution, community detection and graph queries, by Graphify Labs under the Apache-2.0 license. NeuroSquad installs it for you and draws its graph.
Questions about Graphify
See how your agents read your code
Download NeuroSquad, add the Graphify card, and draw an arrow.