Supermodel CLI

Save 40%+ on agent token costs with code graphs.

Supermodel maps every file, function, and call relationship in your repo and writes a .graph file next to each source file. Your agent reads them automatically via grep and cat. No prompt changes. No extra context windows. No new tools to learn.

📖 Full CLI docs: docs.supermodeltools.com/cli/install

Linux / Mac

curl -fsSL https://supermodeltools.com/install | sh

npm (cross-platform)

npm install -g @supermodeltools/cli

How it works

1. Map your codebase

supermodel

Uploads your repo to the Supermodel API, builds a full call graph, and writes .graph shard files next to every source file. Stays running to keep files updated as you code.

2. Your agent reads the graph automatically

.graph files are plain text. Any agent that can read files — Claude Code, Cursor, Copilot, Windsurf — picks them up automatically through its native search & file reading tools. No configuration needed on the agent side.

3. Ask anything

Your agent now has full visibility into your call graph, imports, domains, and blast radius — for every file in the repo, not just the ones open in the editor.

Works with any AI agent

.graph files are plain text read via grep and cat. There is no agent-specific integration required.

For live updates in Claude Code, add this hook to .claude/settings.json:

{
  "hooks": {
    "PostToolUse": [{
      "matcher": "Write|Edit",
      "hooks": [{"type": "command", "command": "supermodel hook"}]
    }]
  }
}

The supermodel setup wizard installs this automatically if Claude Code is detected.

Installation

macOS

brew install supermodeltools/tap/supermodel

Linux / macOS (curl)

curl -fsSL https://supermodeltools.com/install | sh

On first run, supermodel launches the setup wizard automatically when attached to a terminal.

From source

git clone https://github.com/supermodeltools/cli
cd cli
go build -o supermodel .

Quick start

cd your/repo
supermodel                # first run: launches setup wizard, then watches for changes

Commands

File mode

Manages .graph.* shards written next to each source file. Agents read these without making API calls.

Tell your agent about graph files

Add the Supermodel graph-file prompt to CLAUDE.md, AGENTS.md, or your agent's instruction file:

supermodel skill >> CLAUDE.md

Or manually add:

This repository has .graph.* files next to source files containing code relationship data from Supermodel.
For src/Foo.py, the graph file is src/Foo.graph.py.
Each .graph file can include [deps], [calls], and [impact] sections.
Read the .graph file before the source file to understand dependencies, call relationships, and blast radius before making changes.

On-demand analysis

Code tools

Agent integration

Auth and config

Add a badge to your README

[![Supermodel](https://img.shields.io/badge/supermodel-enabled-blueviolet)](https://supermodeltools.com)

Configuration

Save 40%+ on agent token costs — start free → supermodeltools.com/trial

Settings are stored at ~/.supermodel/config.yaml. Environment variables override file values.

api_key: smsk_...        # or SUPERMODEL_API_KEY
api_base: https://...    # or SUPERMODEL_API_BASE (default: https://api.supermodeltools.com)
output: human            # human | json
files: true              # set false to disable .graph.* writing globally (or SUPERMODEL_FILES=false)

For CI or non-interactive environments:

SUPERMODEL_API_KEY=smsk_live_... supermodel analyze

MCP setup

To expose Supermodel graph tools directly to Claude Code via the Model Context Protocol, add the server to ~/.claude/config.json:

{
  "mcpServers": {
    "supermodel": {
      "command": "supermodel",
      "args": ["mcp"]
    }
  }
}

Exposed MCP tools: analyze, dead_code, blast_radius, get_graph.

Links

Questions? Open an issue or email abe@supermodel.software.