Skip to content

Stop babysitting
your coding agents.

Run code reviews and bug fixes in the cloud. aictrl follows the workflow you define and uses your team’s skills. You can inspect each step and its result.

Start free Watch product demo

20 free executions per organization. Models included. No card.

aictrl.dev/org/aictrl/workflows/pr-review-fix
Workflow canvas for PR Review then Fix, managed in code from the repository: a label trigger, a review step and a fix step inside a loop of up to two rounds, and an end node.The run of that workflow, titled by the pull request it ran for and linked to it, marked completed, with the review and fix steps completed inside the two-round loop of the execution graph.The review step of that run, opened from the execution graph: the agent activity leads with its Completed status, 212 tool calls and the agent's latest update, review complete with all eight prior findings verified, beside Step details.The fix step of the same run: completed after 89 tool calls, with the agent's latest update, the verdict for pull request 5880: 10 fixed, 1 deferred, fixes committed and pushed, beside the same Step details.

The process is a YAML file in your repo, run step by step as cloud agents.

Workflows as YAML in your repoSteps run in isolated containers on aictrl cloudInstalls as a GitHub app, MIT headless CLI

01 / What teams run first

Four things to run
on day one.

01 Code review

Every pull request gets findings and a verdict.

Severity, file and line for each finding, a Blocked or Ready verdict, and a step-by-step explanation. A person merges.

The pull requests list: each open pull request with its review verdict, Blocked, Pending or Needs re-review, its highest severity and its count of open findings.
AI code review

03 Triage with a person in the loop

Agents do the legwork. People make the calls.

A chat message opens the flow, a person fills in the annotations, and the agent files the issue. The review loop picks the change up from the pull request.

A workflow canvas: a Telegram chat-message trigger, a user-input step that collects UI annotations in three fields, an agent step that creates the front-end issue, and the end node.
Approval gates and inputs

04 MCP

Run it from the agent you already use.

Claude Code, Codex, Cursor or ChatGPT start reviews, read findings and record verdicts through the aictrl MCP server.

The MCP tool reference: list_tasks, start_task, get_task_execution, list_pr_findings, record_finding_verdict and query_context, with what each does and whether it writes.
Code review over MCP

02 / Start small. Build on it.

Your next review
is a good place
to start.

Try the open engineering skills in your own agent. Connect aictrl when you're ready for shared context, controlled execution, and team visibility.

Read the setup guide
Terminal / install skillsOpen source
npx skills add aictrl-dev/skills

Use the public skills locally without an aictrl account.

  1. 01

    Install your playbooks

    Bring reusable skills into your preferred coding agent.

  2. 02

    Put one workflow to work

    Start with a code review, a design review, or a focused implementation.

  3. 03

    Connect your team

    Add shared standards, workflow execution, and evidence through aictrl.

Inspect the skills on GitHub

03 / Before you build

Good questions.
Direct answers.

Go deeper in the docs
How does aictrl work with my coding agent?

Install the free engineering skills in Claude Code, Codex, OpenCode, Cursor, or Gemini CLI. When you need shared team context or workflow execution, connect your agent to https://aictrl.dev/mcp through MCP and OAuth. See the plugins and setup options.

What connects the four pillars?

Skills define how your team works. Execution composes those skills into workflows. Observability captures traces and evidence from the work. Knowledge supplies codebase context across the system. Explore the engineering lifecycle.

Can I try it before connecting my team?

Yes. Use the public skills locally without an aictrl account. You can also start aictrl for free, with no credit card required. See the plans and their limits.

Where can I review security and data handling?

Read our privacy policy for data handling details and the documentation for connection and authorization setup. For local protection of sensitive data before model requests, explore hush, the plugin gateway.

Build with intelligence. Ship with confidence.

Make great engineering
your agents' default.

Get started See aictrl in action