العربية

Claude Code workflow

Graphify for Claude Code: persistent project context without re-reading everything

Graphify turns a codebase into a structured knowledge graph that Claude Code can use as a project map. The goal is simple: spend less context rediscovering the repository and more context on the change you actually want to make.

Important: this is not literally infinite memory, and there is no fixed token-saving percentage. The benefit depends on the project, the task and how much repeated repository discovery the graph replaces.

What Graphify changes

Persistent project map

Files, functions, classes and dependencies become connected project knowledge instead of isolated snippets.

Faster orientation

A coding agent can start from relevant relationships instead of repeatedly scanning large parts of the repository.

Impact analysis

The graph can help answer questions such as what depends on a file or what may be affected by a change.

Visual inspection

The generated graph can be opened in Obsidian Graph View so you can inspect project relationships visually.

The workflow I use

1

Run Graphify on the project

Generate the project knowledge graph from the repository you want Claude Code to work on.

2

Keep the graph with the project

Graphify creates project output that can persist between coding sessions instead of starting from zero every time.

3

Open the graph in Obsidian

Open the graphify-out/obsidian folder as an Obsidian vault when you want to inspect the relationships visually.

4

Let Claude query the map

Use the graph to narrow the context Claude needs before editing, debugging or tracing dependencies.

Why this can save context and tokens

Large coding sessions often waste context on the same job: locating files, tracing imports and reconstructing how parts of a project connect. A persistent graph can reduce that repeated discovery. That does not mean token use becomes zero; it means more of the available context can be spent on implementation, debugging and reasoning instead of re-learning the repository.

Less context for rediscovery. More context for the actual work.

Where it is most useful

  • Large or unfamiliar repositories
  • Projects you return to across many Claude Code sessions
  • Refactors where dependency and impact analysis matter
  • Multi-agent workflows where several coding agents need the same project map
  • Codebases where you want a human-readable visual graph in Obsidian

Graphify + Claude Code FAQ

Does Graphify give Claude Code infinite memory?

Not literally. A better description is persistent project memory: a reusable map of the codebase that can survive beyond one context window.

Does Graphify reduce token usage?

It can reduce repeated context usage when Claude would otherwise need to rediscover the same repository relationships. The amount varies, so I do not use a fixed percentage or multiplier.

Is Obsidian required?

Obsidian is useful for visualizing the generated graph. The core idea is the persistent project knowledge itself; Graph View is the human-friendly way to inspect it.

Is this only useful for huge projects?

No, but the value usually grows as the repository becomes larger, more connected or more frequently revisited.

What I am testing next

I am testing this as part of a broader workflow for building websites and AI-assisted projects with Claude Code. I publish the useful parts, the limitations and the workflows that are worth keeping.