# Getting Started with Synlets: Setup Guide & Best Practices

From signup to your first AI-generated pull request, plus CI/CD best practices to safely integrate AI agents into your workflow.

Synlets Team

Developer Experience

January 3, 2026

7 min read

## Getting Started with Synlets: Setup Guide & Best Practices

Ready to add AI teammates to your workflow? This guide walks you through setting up your first project, getting your first AI-generated pull request, and configuring your CI/CD pipeline to safely validate agent work.

## Step 1: Create Your Account

Head to synlets.com and sign up. You can use Google, GitHub, or email — whatever's fastest.

## Step 2: Connect Your Integrations

Once you're in, head to **Integrations** in the sidebar:

- **Source control** — Connect GitHub or GitLab
- **Task management** — Connect Jira, Asana, or your preferred tool

Select which repositories and boards Synlets should have access to.

> **Tip**: Start with a non-critical repo to test the workflow.

## Step 3: Create a Project

This is where it all comes together. A **Project** links your task management with your repositories. Head to **Projects** and create one:

1. **Name your project** — e.g. "Backend API" or "Mobile App"
2. **Connect repositories** — Select which repos this project covers
3. **Configure task management** — Link your Jira board or Asana project
4. **Set up statuses** — Map your workflow statuses so the agent knows where tasks are in the pipeline:

| Status | What it means |
| --- | --- |
| **Todo** | Task is ready to be picked up |
| **In Progress** | Agent is actively working on it |
| **In Review** | PR created, waiting for human review |
| **Merged** | PR approved and merged — task complete |
| **Rejected** | PR was rejected — agent can reattempt or you reassign |

5. **Set a label** — Tasks are picked up based on labels. Add a label like `synlets` or `ai-agent` to any ticket you want the agent to work on

## Step 4: Add Knowledge (Optional but Recommended)

Upload your team's knowledge base to help the agent understand your standards:

- **Confluence pages** — Architecture docs, coding guidelines, API specs
- **Notion docs** — Team conventions, style guides, decision records
- **Any documentation** — Anything you'd share with a new hire

The more context you provide, the better the agent's output. Think of it like onboarding a new team member — the agent reads everything before writing a single line of code.

## Step 5: Assign Your First Task

Label a ticket in your task manager with your configured label. The agent will automatically pick it up and:

1. Read the ticket requirements
2. Explore your codebase for context
3. Reference your knowledge base docs
4. Implement the solution
5. Create a pull request with a clear description

You don't need to watch or prompt. Assign it and come back to a finished PR.

## Step 6: Review and Merge

Check your GitHub or GitLab for the new PR. Review the changes like you would any teammate's code:

- Code follows your conventions
- Commit messages are clean
- PR description explains the changes

If it looks good, merge it. If you have feedback, comment on the PR — the agent will read your comments, make changes, and push updates.

## Best Practice: Let Your CI Pipeline Verify the Work

Here's something important to understand: **Synlets doesn't run your code**. It writes code and creates pull requests — but it never executes, builds, or deploys anything.

This is by design. Your existing CI/CD pipeline is the safety net.

### How to Set It Up

Configure your repository so that PRs automatically trigger your CI pipeline:

1. **Branch pattern triggers** — Most CI systems (GitHub Actions, GitLab CI, CircleCI, etc.) let you trigger pipelines on specific branch patterns. Synlets creates branches using standard conventions like `feat/*`, `bugfix/*`, `task/*`, or `improvement/*` depending on the ticket type. Make sure your CI config runs on these patterns:

```yaml
# Example: GitHub Actions
on:
  pull_request:
    branches: [main, develop]
  push:
    branches:
      - 'feat/**'
      - 'bugfix/**'
      - 'task/**'
```

2. **Run your test suite** — Unit tests, integration tests, type checking, linting. If it's in your pipeline, it runs on agent PRs too. A failing test means the agent's code doesn't pass muster — just like with any developer.

3. **Require checks to pass before merge** — Enable branch protection rules so PRs can't be merged until CI passes. This ensures nothing from the agent (or anyone else) reaches your main branch without passing your quality gates.

### Bonus: Ephemeral Preview Environments

For teams with QA processes, the ideal setup is **ephemeral environments** — temporary deployments that spin up automatically for each PR.

Tools like Kubernetes with ArgoCD, Flux, or custom preview environment controllers — or simpler options like Vercel, Netlify, and Railway — can deploy each PR to a unique URL. Your QA team can then:

- Click the preview link on the PR
- Test the changes in a real environment
- Approve or request changes before merge

This works especially well with Synlets because:

- Agent creates PR → CI runs → Preview deploys → QA tests → You merge
- No one needs to pull code locally or run anything manually
- The entire flow is automated except for the final human review

### The Result

Synlets writes the code. Your CI verifies it works. Your team reviews and merges. At no point does unvalidated code reach production — the same safeguards that protect you from human errors protect you from agent errors.

## Bonus: Set Up AI Code Reviews

Want an extra layer of quality? Enable the **PR Review Agent** on your repositories. When the Project Agent creates a PR, the Review Agent will automatically:

- Review the code changes against your standards
- Flag potential issues or improvements
- Create a child PR with fixes against the branch — not just comments, but actual code changes you can merge

So when you open the PR, the Review Agent has already caught issues and provided fixes in a child PR. Merge the fixes, and by the time you review, the basics are already handled.

## What's Next?

- **Scale up** — Add more labels to more tickets as you build trust
- **Add knowledge** — The more docs you upload, the better the output
- **Check analytics** — Track ACU usage, tasks completed, and PR merge rates
- **Enable PR reviews** — Let AI review human PRs too, not just agent work

Welcome to the future of engineering teams.