Vericore

Getting Started

From a fresh install to repository understanding and a verified code change.

Choose your path

I want to understand a repository

Start with:

vericore analyze /path/to/repository

Then open:

/path/to/repository/output/index.html

For machine-readable deterministic state:

vericore evidence-graph /path/to/repository --json
vericore reality /path/to/repository --json

I want to understand a risky change

Use deterministic repository evidence:

vericore repo-qa "Why is PaymentService risky?" --path /path/to/repository
vericore impact /path/to/repository src/main/kotlin/com/example/PaymentService.kt --json
vericore architecture /path/to/repository --json

I want to let an AI agent use Vericore

Start the local MCP server:

vericore mcp

Then read MCP for the canonical tool surface and trust boundary.

I want to make a code change safely

Prepare the repository before editing:

vericore prepare "add payment validation" --path /path/to/repository

Make the code change, run the project’s normal tests, then verify:

vericore verify --path /path/to/repository

The verification step is bound to the persisted Agent Change Contract created by prepare.

Prerequisites

For a source build, use JDK 21 or newer:

Released platform archives bundle a Java runtime, so a separate JDK is not required for normal end-user use.

1. Install from a release archive

Download the published archive from GitHub Releases.

Platform archives bundle a Java runtime for normal end-user execution:

Windows:      bin\vericore.bat --version
Linux/macOS:  ./bin/vericore --version

2. Install with one command

Linux x64 / macOS

curl -fsSL https://raw.githubusercontent.com/sonii-shivansh/Vericore/main/scripts/install.sh | bash

Windows PowerShell (x64)

irm https://raw.githubusercontent.com/sonii-shivansh/Vericore/main/scripts/install.ps1 | iex

The installer downloads the latest published platform archive, verifies its SHA-256 checksum from SHA256SUMS, installs the bundled Java runtime, and avoids requiring a separate JDK.

Supported targets: Linux x64, macOS x64, macOS arm64, and Windows x64. After installation, open a new shell if PATH was updated.

3. Build from source

If you are developing Vericore itself:

git clone https://github.com/sonii-shivansh/Vericore.git
cd Vericore
./gradlew --no-daemon clean test
./gradlew --no-daemon installDist

The installed CLI is:

build/install/vericore/bin/vericore

Verify it:

./build/install/vericore/bin/vericore --version
./build/install/vericore/bin/vericore --help

For every command, option, default, output artifact, and failure behavior, use the CLI Reference.

4. Check the local installation

Run Doctor immediately after installation; you do not need to be inside a Git repository:

vericore doctor

For a specific repository, provide its path:

vericore doctor --path /path/to/repository

Doctor checks the Java runtime, diagnostic path, Git repository context when available, effective AI configuration, credentials when configured, and Gemini reachability when applicable. Being outside a Git repository is a warning—not a failure—because Doctor is also an installation and configuration diagnostic. Missing AI credentials are a warning rather than a failure because deterministic repository analysis does not require AI.

5. Go deeper

Goal Read
Exact command syntax CLI Reference
Understand deterministic repository identity Engineering Reality
Understand safe changes Change Safety
Understand the system Architecture
Integrate an AI agent MCP
Integrate REST API
Review data handling Data & Privacy
Contribute code Development

Troubleshooting principle

Start with the exact command and sanitized output. For CI-only failures, inspect the corresponding GitHub Actions job and artifact rather than inferring from a local result.