Quickstart

Get from zero to your first AI-driven pull request in minutes.

1. Prerequisites

• Node.js 22+
• Claude CLI installed and authenticated
• GitHub CLI (gh) installed and authenticated
• Git configured with push access to your repo

2. Install

git clone https://github.com/<your-org>/reygent.git
cd reygent
npm install
npm run build
npm link

Verify it works:

reygent --version

3. Configure API Keys (Optional)

Only needed if you want to pull specs from an issue tracker instead of a markdown file.

Create a .env file in your target project root:

# Linear
LINEAR_API_KEY=lin_api_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

# Jira
JIRA_URL=https://your-company.atlassian.net
JIRA_EMAIL=you@company.com
JIRA_API_TOKEN=xxxxxxxxxxxxxxxxxxxxxxxx

4. Initialize Your Project

Navigate to the repo you want reygent to work on:

cd /path/to/your-project
reygent init

This creates .reygent/config.json with default agent definitions. You can skip this step — reygent falls back to built-in agents.

5. Write a Spec

Create a spec.md file describing what you want built:

# Add health-check endpoint

## Requirements
- GET /health returns 200 with JSON body { "status": "ok" }
- Responds in under 50ms
- No authentication required

## Acceptance Criteria
- Endpoint is reachable at /health
- Returns correct JSON response
- Unit test covers success case

Or generate one from a description:

reygent generate-spec "Add a health-check endpoint" --output spec.md

Or point directly at a Linear or Jira issue:

reygent run --spec ENG-123
reygent run --spec https://linear.app/your-team/issue/ENG-123/your-issue

Issue keys (e.g. ENG-123) are resolved based on which credentials are configured in .env. If both Linear and Jira are configured, Linear is tried first.

6. Run the Workflow

reygent run --spec spec.md

Reygent runs 7 stages automatically:

Plan → Implement → Unit Tests → Functional Tests → Security Review → PR Create → PR Review

You'll be prompted to choose:

• Auto-approve mode — agents run without asking permission for each file edit (faster, runs dev + QE in parallel)
• Clarification preference — whether the planner asks you questions or makes assumptions

Fully autonomous mode

reygent run --spec spec.md --auto-approve --skip-clarification

Preview without executing

reygent run --spec spec.md --dry-run

7. Customize Config

Use the interactive config command to set your provider, model, and per-agent overrides:

reygent config

This walks you through selecting a provider and model, then lets you customize each agent individually. See the config command reference for details.

You can also edit .reygent/config.json directly:

{
  "agents": [
    {
      "name": "dev",
      "description": "Write, edit, and refactor implementation code",
      "systemPrompt": "You are the Dev agent. Follow our team's coding standards...",
      "tools": ["read", "write", "bash", "search"],
      "role": "developer"
    }
  ],
  "model": "claude-sonnet-4-5-20250929"
}

You can also override the model per-run:

reygent run --spec spec.md --model claude-opus-4-6

8. Install Skills from the Registry

Browse and install community skills without manually copying files:

# See what's available
reygent skills list

# Install a skill to your project
reygent skills add code-reviewer

# Use it
reygent agent code-reviewer

# Install globally (shared across all projects)
reygent skills add code-reviewer --global

# Remove when done
reygent skills remove code-reviewer

See Skills for the full guide.

Useful Commands

Command	Description
reygent init	Initialize .reygent/ config in current project
reygent generate-spec "..."	Generate spec from description
reygent spec spec.md	Load and display a parsed spec
reygent agent dev	Start interactive session with an agent
reygent agent dev --spec spec.md	Interactive session with spec context
reygent run --spec spec.md	Run full 7-stage workflow
reygent skills list	Browse available skills in the registry
reygent skills add <name>	Install a skill from the registry
reygent skills remove <name>	Remove an installed skill

Next Steps

• README — full documentation, all options, agent details
• Commands Reference — every command and flag
• Agents Guide — how agents work and how to customize them
• Workflows — visual diagrams of the workflow and retry logic
• Skills — extend reygent with custom skills
• Architecture — technical deep-dive