How to use Graphify in Perplexity Computer
By Steven Van ·
Perplexity Computer users can map a public GitHub repo with the Graphify skill, free through October 31, 2026, up to 5,000 credits each.
Graphify now has a step-by-step guide for running it inside Perplexity Computer. The setup is to install the Graphify skill from Plugins, type /graphify in a Computer task, and paste a link to a public GitHub repository. Computer clones the code and builds a knowledge graph. The guide comes from the Graphify team.
Installing the skill
- Open Plugins in the left sidebar and go to the Skills tab.
- Search for graphify and open Graphify, listed under Developer.
- Select Install. The button becomes an Enabled toggle.
Perplexity is making tasks that use /graphify free through October 31, 2026, up to 5,000 credits per Computer user.
Starting a task
Select New, set the mode to Computer, type /graphify and choose Use to pick the skill, then name the repository. The example prompt was "map the codebase at https://github.com/Graphify-Labs/graphify". Computer cloned the public repository on its own, with no connector, upload or sign-in. It picked 102 code files and three docs and finished in about five minutes.
What comes back
Each run produces three artifacts in the task:
- Graphify Codebase Map, an interactive app that opens in the task.
- Graphify Codebase Map Bundle, a ZIP with graph.json, graph.html, the raw extraction and a notes file.
- Graphify Codebase Graph Report, covering key functions, the main flow and questions worth asking.
For the Graphify repository, the map held 4,022 nodes, 10,079 directed edges and 192 communities.
Exploring the map
The map is a force-directed graph. Each edge keeps its direction, relation type, a confidence tag and its source file. Search nodes finds a file or function by name. Clicking a node opens its type, community, source path, degree and full neighbor list. For example, extract.py shows a degree of 442 in the community Extract Core & Template Masking. The Communities panel lists each cluster with its size and lets you show or hide it, and Light, Dark and System switch the theme.
The map stays in the task, so follow-up questions answer from the graph rather than rereading files. Suggested questions include which functions call the authentication middleware and how a request gets from the API router to the database.
Running it locally
The open source engine needs Python 3.10 or newer and uv. Run uv tool install graphifyy, then graphify install, which detects the assistant and registers the skill. Typing /graphify . in Claude Code, Cursor, Gemini CLI or another supported assistant produces graph.html, GRAPH_REPORT.md and graph.json in graphify-out/.
