# Your AI built it. Buddy ships it.

Buddy is CI/CD for apps built with Lovable, Bolt, Cursor or Claude Code. Every push is tested, every branch gets a preview link, and nothing goes live without your OK.

_Last updated: 2026-11-09_

Source: https://buddy.works/ci-cd/

[Start free](https://buddy.works/sign-up)

## Three common gaps. Buddy closes each one.

### Secrets stuck in the builder

Keep them encrypted, per environment.

```bash
bdy var add DATABASE_URL --file - --encrypted --environment production
```

[Variables in the CLI](https://buddy.works/docs/cli/variable/create)

### A different database in production

Test on real PostgreSQL, MySQL or Redis.

```bash
services: - type: "MYSQL"
```

[Services and databases](https://buddy.works/docs/actions/services-and-databases/predefined)

### Nobody checked before going live

Deploys wait for your OK.

```bash
type: "WAIT_FOR_APPLY"
```

[Manual approval](https://buddy.works/docs/pipelines/parameterized-pipelines/manual-approval)

## From push to production

1. **Push** - You or your agent push to Git.
2. **Build and test** - On every push, with a real database.
3. **Preview** - A live link for the branch.
4. **Review** - A teammate or an AI agent checks it.
5. **Approve** - A person says go.
6. **Deploy** - To your server, cloud or cluster.

## Let your agent set it up

Using Claude Code, Codex or Cursor? Your agent can write and run the pipeline for you, from 150+ ready-made actions.

### One command: /ci

Install the Buddy plugin and type /ci. It writes buddy.yml, creates the pipeline and runs it. Needs the Buddy CLI.

```bash
claude plugin marketplace add buddy/buddy-plugin
claude plugin install buddy@buddy-plugin
# then, inside Claude Code:
/ci
```

[Buddy plugin for Claude Code](https://buddy.works/docs/ai-agents/agents/claude)

### Add the Buddy skill

Install the Buddy CLI and skill, link the project and start Codex. Then ask it to build and deploy.

```bash
npm install -g bdy
bdy login
bdy proj link
npx skills add buddy/buddy-plugin
codex
```

[Buddy for Codex](https://buddy.works/docs/ai-agents/agents/codex)

### Add the Buddy skill

Same setup, then start the Cursor agent (⌘I / Ctrl+I) or the Cursor CLI and ask it to ship.

```bash
npm install -g bdy
bdy login
bdy proj link
npx skills add buddy/buddy-plugin
cursor-agent
```

[Buddy for Cursor](https://buddy.works/docs/ai-agents/agents/cursor)

### Any MCP agent

Add the Buddy MCP server. Your agent can then run pipelines, read logs and start previews.

```json
{
  "mcpServers": {
    "buddy": {
      "url": "https://mcp.buddy.works/mcp"
    }
  }
}
```

[Buddy MCP server](https://buddy.works/docs/ai-agents/mcp)

### Or do it yourself

Start and follow pipelines from your terminal.

```bash
npm install -g bdy
bdy login
bdy pipeline run start deploy -b main
```

[Buddy CLI](https://buddy.works/docs/cli/getting-started)

## Safety checks for AI-written code

Checks in the pipeline can't be skipped. Every change gets the same tests, an AI review and a person's approval before it reaches production.

### AI review on every pull request

Claude Code, Codex or Cursor review each pull request inside the pipeline.

```yaml
- pipeline: on-pull-request
  events:
  - type: PULL_REQUEST
    branches:
    - master
  actions:
  - action: Claude Code
    type: CLAUDE_CODE
    integration: anthropic
    prompts:
    - make pull request review
    model: sonnet
    claude_args: --max-turns 3
    dangerously_skip_permissions: true
```

[AI agents in pipelines](https://buddy.works/docs/ai-agents/pipelines)

### Test against a real database

Attach a database to the build. It runs right next to your tests.

```yaml
- action: "Run tests"
  type: "BUILD"
  docker_image_name: "library/node"
  docker_image_tag: "22"
  commands: |-
    npm ci
    npm test
  services:
    - type: "MYSQL"
      version: "5.7"
```

[Build action in YAML](https://buddy.works/docs/yaml/yaml-actions/build)

### Ten commits, one build

A new push cancels the older runs on the same branch.

```yaml
- pipeline: "Run tests"
  events:
    - type: PUSH
      refs:
        - "refs/heads/*"
  concurrent_pipeline_runs: true
  skip_queued_runs: SAME_REF
  cancel_inprogress_runs: SAME_REF
  actions:
  - action: "Run test suite"
    type: "BUILD"
    docker_image_name: "node"
    docker_image_tag: "22"
    commands: |-
      npm ci
      npm test
```

[Concurrency and parallelism](https://buddy.works/docs/pipelines/concurrency-and-parallelism)

### A person decides what ships

The run pauses and asks: deploy to production? Only the people you choose can say yes.

```yaml
- action: "Wait for apply"
  type: "WAIT_FOR_APPLY"
  trigger_time: "ON_EVERY_EXECUTION"
  comment: "Do you want to deploy ${BUDDY_RUN_COMMIT} to Production?"
```

[Wait for apply in YAML](https://buddy.works/docs/yaml/yaml-actions/wait-for-apply)

## When agents open your pull requests

### A pipeline for every pull request

Runs start when a pull request opens or changes - in GitHub, GitLab or Buddy Git.

[Git event triggers](https://buddy.works/docs/pipelines/introduction/triggering-pipelines/on-git-event)

### Results on the pull request

The run status shows up right on the pull request.

[Pull request testing](https://buddy.works/docs/pipelines/pipeline-examples/pull-request-testing)

### See who started each run

Runs from the API, the CLI or MCP show who triggered them.

[Pipeline notes](https://buddy.works/docs/pipelines/introduction/pipeline-notes)

## Everything else you need

### Fast, cached builds

Dependencies and Docker layers are reused between runs.

[Cache](https://buddy.works/docs/pipelines/cache)

### Monorepo ready

Build only the app whose files changed.

[Conditional runs](https://buddy.works/docs/pipelines/conditional-executions)

### Parallel runs

Split tests and run version combinations side by side.

[Pipeline loop](https://buddy.works/docs/pipelines/introduction/pipeline-loop)

### UI or YAML

Build it by clicking, keep it as buddy.yml. Both stay in sync.

[YAML and GUI](https://buddy.works/docs/yaml/yaml-pipelines/yaml-gui)

### Deploy anywhere

Your server, AWS, Google Cloud, Azure, DigitalOcean, Netlify or Kubernetes.

[Deployments](https://buddy.works/docs/basics/deployments)

### No stored cloud keys

OIDC gives each action a short-lived token for AWS, Google Cloud or Azure.

[OIDC](https://buddy.works/docs/integrations/oidc)

## Questions

### How do I deploy a Lovable or Bolt app to production?

Push it to GitHub or GitLab and connect the repository to Buddy. Every push is built and tested, each branch gets a preview, and production deploys after your approval.

### Why does my app work in the builder's preview but not in production?

The preview had env vars or a database that production doesn't. Buddy keeps variables per environment and tests every push against a real database, so the build fails before your users notice.

### How do I set environment variables and secrets for production?

Add them as Buddy variables for each environment. Mark secrets as encrypted - they stay hidden in logs and in the API. For AWS, Google Cloud and Azure, use OIDC instead of stored keys.

### How do I get a preview environment for every branch?

Add a pipeline that creates a preview Sandbox for each new branch, and one that removes it when the branch is deleted. Each preview has its own link, which you can keep private.

### Can Claude Code or another AI agent set up my CI/CD pipeline?

Yes. In Claude Code, install the Buddy plugin and run /ci - it writes buddy.yml, creates the pipeline and runs it. Other agents connect through the Buddy MCP server.

### How do I check AI-generated code before it goes to production?

Test every push, add an AI review to pull requests, and require a person's approval before production.

### Is Buddy an alternative to GitHub Actions?

Yes. Buddy works with your GitHub repository, runs on pushes and pull requests, and adds previews, environments and approvals in one place.

### How do I set up CI/CD for AI coding agents?

Run a pipeline on every pull request, cancel outdated runs, keep production behind an approval, and connect agents through MCP or the CLI so they can read the results.

### Does Buddy work with monorepos?

Yes. Each app gets its own pipeline that runs only when its files change.

## Built it? Ship it today.

Connect a repo and run your first pipeline.

[Start free](https://buddy.works/sign-up)
