AI coding agents like OpenCode, Claude Code, and Cursor have completely transformed how developers work. But running them requires a local environment, a terminal, and usually a human at the keyboard.

Codecloud is an API for running AI coding agents in the cloud, based on the full power of OpenCode. You connect your GitHub repository, make a POST request with a prompt, and the agent clones your repo, implements the changes, and opens a pull request. No local setup, and no infrastructure or state to manage.

Unlike running agents in a docker container inside CI / CD, codecloud API calls start the agent in less than a second and don't take up build hours from your CI pipeline.

Why an API?

APIs always unlock automation. Instead of manually running agents, you can trigger runs programmatically. A webhook fires, a ticket gets created, a cron job runs—and code gets written.

This opens up use cases that you can use to automate your entire product development process:

Issue tracker integrations: Connect Linear, Jira, GitHub Issues, Asana, or ClickUp. When a ticket is created, an agent reads it and opens a PR.
Slack and Discord bots: Let team members request code changes by messaging a bot. The agent handles implementation.
No-code automation: Build workflows in n8n, Zapier, Make (Integromat), or Pipedream that write code without any manual intervention.
Custom internal tools: Add AI coding to your internal platforms. Let non-developers request features or fixes.

How it works

The workflow is simple:

Connect GitHub: Authorize the codecloud GitHub app to access your repositories
Make an API call: Send a POST request with your repo, prompt, and preferred model (Claude Opus 4.5, GPT-5.2, Gemini 3, or others).
Agent runs: We spin up a sandboxed environment, clone your repo, and run the agent with your prompt.
Get results: The agent opens a PR, or you poll for completion, or we send a webhook to your endpoint.

curl -X POST https://codecloud.dev/api/v1/agents \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "repo": "your-org/your-repo",
    "prompt": "Add input validation to the signup form",
    "model": "claude-sonnet-4-5",
    "auto_create_pr": true
  }'

The response includes a run ID. You can poll for status, stream logs, or just wait for the PR to appear on GitHub.

Supported integrations

Because codecloud is a standard REST API, it works with any tool that can make HTTP requests. Some popular integrations:

Issue trackers: Linear, Jira, GitHub Issues, Asana, Monday.com, Notion, Trello, ClickUp, Shortcut, YouTrack, Azure DevOps

Automation platforms: n8n, Zapier, Make, Pipedream

Communication tools: Slack, Discord, Microsoft Teams, Telegram

CI/CD: GitHub Actions, GitLab CI, CircleCI, Jenkins, Buildkite, Travis CI, Azure Pipelines

Monitoring: Sentry, Datadog, PagerDuty, Opsgenie — trigger bug fixes when errors spike

Use cases

Automated bug fixing

Connect your issue tracker to codecloud. When bugs are reported, an agent reads the description, finds the relevant code, implements a fix, and opens a PR. Your developers review instead of implement.

This works especially well for well-described bugs. "Button doesn't submit form on Safari" with reproduction steps gives the agent enough context to investigate and fix.

Backlog automation

Every team has a backlog of small improvements that never get prioritized. Typos, missing error messages, documentation updates, minor UI tweaks. Set up a workflow that processes these automatically. The agent works through them while your team focuses on larger features.

Implementation plans

Not ready to auto-merge? Use plan mode. The agent analyzes your codebase and returns a detailed implementation plan without making changes. Useful for scoping tickets, onboarding new team members, or validating approaches before committing to them.

Code review bots

Build a bot that reviews incoming PRs. The agent reads the diff, checks for issues, and either approves or opens a follow-up PR with suggested improvements. Integrates with GitHub webhooks.

Product feature: AI coding for your users

Building a dev tool, a website builder, or a no-code platform? Integrate codecloud to give your users AI coding capabilities. They describe what they want, your product calls our API, they get working code.

Any model, your API keys

codecloud supports multiple AI providers. Use Claude, OpenAI, Gemini, or other models supported by OpenCode. Bring your own API keys and pay providers directly.

Different models suit different tasks. Sonnet handles most bugs and features efficiently. Opus excels at complex refactoring and architectural changes. Haiku works for simple, high-volume tasks.

Zero infrastructure

Running coding agents at scale is operationally complex. You need sandboxed environments, compute resources, proper isolation, and queue management. We handle all of that.

Each agent run gets a fresh, isolated environment. Your code is cloned, the agent runs, results are delivered, and the environment is destroyed. No persistent state, no security concerns about code mixing between runs.

Webhooks and real-time updates

Pass a webhook URL in your request and we'll POST to it when the run completes. Useful for updating tickets, posting to Slack, or triggering downstream workflows.

{
  "repo": "your-org/your-repo",
  "prompt": "Fix the memory leak in the cache module",
  ...
  "webhook_url": "https://your-app.com/webhooks/codecloud"
}

The webhook payload includes the run status, any PR URL, agent output, and metadata. You can also poll the status endpoint or stream logs in real-time.

Getting started

Sign up at codecloud.dev, connect your GitHub account, and grab an API key. The documentation covers authentication, request formats, and response handling.

For a practical example, check out our guide on building an n8n workflow that creates PRs from Linear tickets.

Questions or feedback? Use the support portal. For API questions, see the documentation.