Getting started

Install once, then point it at your repo.

This guide covers the first-run flow: create config, set your embedding key, index a codebase, start the MCP server, and optionally run continuous sync.

Install

Pick the install path that fits your machine.

curl

curl -fsSL https://raw.githubusercontent.com/tesh254/raph/main/install.sh | sh

wget

wget -qO- https://raw.githubusercontent.com/tesh254/raph/main/install.sh | sh

PowerShell

irm https://raw.githubusercontent.com/tesh254/raph/main/install.ps1 | iex

Homebrew

brew install --cask tesh254/raph/raph

Installs from the dedicated tap repository at tesh254/homebrew-raph.

Version override
RAPH_REPO=tesh254/raph \
RAPH_VERSION=v0.1.0 \
curl -fsSL https://raw.githubusercontent.com/tesh254/raph/main/install.sh | sh
Release builds can check for newer stable versions automatically. If you want a fully manual setup, see the update controls on the reference page.
Bootstrap

Create the local config and data directories.

The first command sets up the files and folders under ~/.raph/. Then you provide the OpenRouter key used for embeddings.

raph config init
  • ~/.raph/schema.json
  • ~/.raph/raph.json
  • ~/.raph/data/
export OPENROUTER_API_KEY="your-key"

Without an embedding provider, raph falls back to keyword-oriented behavior where possible.

Index

Index a repository and prepare agent config.

Initial scan

raph init --path .

This scans supported files, builds graph relationships, and registers the repository for future background synchronization.

Skip embeddings

raph init --path . --no-embeddings

Use this if you want a faster local pass or do not want remote embedding calls during indexing.

Install MCP entries

raph agents mcp setup --path .

This writes the raph MCP entry for supported agents — globally (user-level config, the default) or per project with --scope local. Interactive terminals are prompted for the scope.

Start the MCP server

raph start

Use this when you want a coding agent to talk to raph over stdio.

Continuous sync

Keep the indexed repository up to date.

The sync worker performs an initial index, tracks supported file changes, updates changed nodes, and removes graph data for deleted files.

raph sync --path .

Starts one detached background worker after registering the repository path.

raph sync --status
raph sync --remove --path .
raph sync --stop

Unregistering a repository removes its graph data by default. Pass --keep-data if you only want to stop tracking it.