# aictrl

> AI workflow orchestration platform for engineering teams, with shared skills, cloud execution, observability, and a codebase knowledge graph.

aictrl (aictrl.dev) is an AI workflow orchestration platform for engineering teams. It gives coding agents reusable skills, isolated cloud execution, evidence and trace capture, and codebase context. The company also publishes eleven free, vendor-neutral engineering skills under an MIT license that work locally without an aictrl account.

## What It Is

aictrl sits between a team's coding agents and its engineering process. Teams define standards as versioned skills, compose them into workflows, run them in isolated cloud containers, and inspect the resulting evidence. Its four pillars are Skills, Execution, Observability, and Knowledge.

It works with agents the team already uses: Claude Code, Codex, OpenCode, Cursor, and Gemini CLI. Connected features use a hosted MCP server at aictrl.dev/mcp with OAuth.

## How the Workflow Works

Workflows are declarative YAML files committed under `.aictrl/workflows/` and applied through git-sync after a pull request is merged. Files that fail schema validation or reference unknown skills are rejected whole. Triggers can start a workflow from GitHub events such as a label being added, a pull request opened or marked ready, a PR comment command, or a Telegram chat message. Each task runs in an isolated cloud container and produces inspectable artifacts such as screenshots, test outputs, commit hashes, and logs.

## Code Review and Knowledge Graph

AI code review tracks findings, review duration, acceptance rate, and noisy rules on pull requests. Connected GitHub repositories can be built into a Knowledge Graph that agents query through `query_context`. Stack Configuration maps repository files to architecture layers, each with specialist agents, skills, and checks. PR Explanations give a plain-language walkthrough of a change.

## Open Engineering Skills and hush

The public skills repository covers issue and bug creation, spec review, implementing code changes, PR creation, code review, root-cause analysis, workflow authoring and execution, design review, UX testing, and more. They install with `npx skills add aictrl-dev/skills` or as Claude Code, Codex, or OpenCode plugins. aictrl also describes hush, a local plugin gateway that replaces sensitive customer data with safe tokens before requests reach model providers and restores the values locally in responses.

## Features
- Versioned skill library shared across agents
- Workflow-as-code in YAML with GitHub event triggers
- Isolated cloud container task runs
- Evidence capture and execution traces
- Execution analytics
- AI code review on pull requests
- PR explanations
- Codebase Knowledge Graph and codebase chat
- Stack configuration by architecture layer
- hush plugin gateway for redacting sensitive data
- Hosted MCP server with OAuth
- Eleven free open-source engineering skills

## Integrations
Claude Code, Codex, OpenCode, Cursor, Gemini CLI, GitHub, MCP, Telegram

## Platforms
WINDOWS, WEB, API, CLI

## Pricing
Subscription-based

## Version
v1.0.2

## Links
- Website: https://www.aictrl.dev
- Documentation: https://aictrl.dev/docs
- Repository: https://github.com/aictrl-dev/skills
- EveryDev.ai: https://www.everydev.ai/tools/aictrl