Skip to content
The N8N of AI Agents

Orchestrate

Build, deploy, and manage collaborative AI agent teams for any purpose. Engineering, marketing, finance, operations. Define the mission, and AgentCrew handles the orchestration.

agentcrew.sh
AgentCrew
Teams
Schedules
Webhooks
Post-Actions
Variables

Settings

Profile
Organization
LinkedIn Content Agent 1 agent

Generates LinkedIn posts with images and Slack alerts

L
Freepik MCP Slack MCP
Survey Moderator 1 agent

Moderates survey responses and exports Excel reports

L
PostgreSQL MCP
AI QA Engineer 1 agent

Reviews code on every Merge Request automatically

L
GitLab MCP
Open Source Docker & Kubernetes Any use case
About

What is AgentCrew

AgentCrew is an open-source AI agent orchestrator. Just as N8N automates workflows regardless of the services involved, AgentCrew orchestrates AI agents regardless of their purpose. Define your team, describe their roles in plain text, deploy, and let them collaborate autonomously.

Purpose-Agnostic

Build teams for software engineering, marketing, finance, research, operations, or any domain. Agent specialization is driven by documentation, not code.

Documentation-Driven

Agents are defined by Markdown files with instructions, roles, and skills. Write what an agent should do, and it becomes that specialist. No code required.

Any Infrastructure

Deploy on Docker for local development or scale to Kubernetes in production. Your infrastructure, your rules.

Capabilities

One platform, any mission

AgentCrew handles the orchestration: team creation, deployment, communication, and monitoring. You focus on defining what your agents should accomplish.

Create Teams for Anything

Build agent teams for any domain: engineering, content creation, data analysis, customer support, legal review. Write Markdown instructions and deploy.

Leader + Workers

Define a leader agent that coordinates tasks and specialized workers that execute them. Like a real team, each agent has a clear role and expertise.

Scheduled Tasks

Automate recurring work with cron-based schedules. Define a prompt, pick a frequency, and AgentCrew deploys the team, executes the task, and records the results automatically.

Webhook Triggers

Trigger agent teams from external systems via HTTP webhooks. Use fire-and-forget or wait-for-response modes with secure token authentication.

MCP Servers

Connect agents to external tools via the Model Context Protocol. Databases, APIs, Kubernetes clusters, and more. Configure once, shared across the entire team.

Post-Actions

Reusable HTTP callbacks that fire automatically after webhook or schedule runs. Notify Slack, PagerDuty, or any API about execution results with template variables and retry logic.

Real-World Examples

See it in action

Real examples of how teams use AgentCrew to automate work and save hours every week.

Automated LinkedIn Content

An AI agent with deep knowledge of your company brand voice and goals. Using AgentCrew Schedules, it runs every Monday, Tuesday, and Friday at 9 AM — scanning industry news, evaluating content opportunities, and crafting LinkedIn posts. It generates custom images via the Freepik MCP and notifies your team on Slack when each post is ready for review.

Agents

Content Strategist (with Freepik MCP + Slack MCP)

Trigger

Schedule (Mon, Tue, Fri at 09:00)

Result

3 ready-to-publish LinkedIn posts per week with custom images and Slack notifications

Survey Response Moderator

An AI agent connected to your database via the PostgreSQL MCP. Through a daily AgentCrew Schedule, it analyzes all new survey responses — flagging inappropriate content, spam, and policy violations based on your business rules. It generates two Excel reports: flagged users with reasons, and validated clean responses.

Agents

Content Moderator (with PostgreSQL MCP)

Trigger

Daily schedule (every day at configured time)

Result

Daily moderation reports: flagged users + validated responses in Excel

AI QA Engineer

Triggered automatically when a Merge Request is opened in GitLab. A CI/CD pipeline calls the AgentCrew Webhook, invoking the agent to perform a comprehensive code review — checking quality, test coverage, duplicate code, security issues, and best practices. The agent posts its review as a comment directly on the Merge Request.

Agents

QA Engineer (with GitLab MCP)

Trigger

Webhook (triggered by GitLab CI on MR open)

Result

Automated code review comment on every MR within minutes

Process

How it works

From zero to a running AI agent team in minutes. No complex setup required.

01

Define your team

Name your team, choose a leader, and add workers. Each agent gets a role, instructions, and optional skills, all defined in plain text.

02

Write the mission

Describe what each agent should do in Markdown. A marketing strategist, a data analyst, a code reviewer. Specialization comes from documentation.

03

Deploy

Click deploy. AgentCrew provisions a Docker network, NATS message bus, and starts the agent container automatically.

04

Chat & Monitor

Send tasks to your team through the chat interface. The leader delegates, workers execute, and you watch it all happen in real time.

Technology

Built with modern tech

Production-ready technologies chosen for performance, reliability, and developer experience.

React
TypeScript
Go
Fiber
Docker
Kubernetes
NATS
SQLite
Claude Code
Astro
Under the Hood

System Architecture

A clean, event-driven architecture where agents communicate asynchronously through a central message bus.

Isolated Containers

Each agent runs in its own container with isolated file system and tools.

Full Observability

Stream logs, agent states, and task progress via WebSocket in real time.

Async by Design

NATS enables non-blocking, pub/sub communication between all agents.

Contact

Let's talk

Have questions about AgentCrew? Want to discuss how AI agent teams can help your organization? Get in touch.

Ready to deploy

Build your first
agent team today

Open source, self-hosted, and running in one command. Define the mission, and AgentCrew handles the rest.

$ curl -fsSL https://agentcrew.sh/install.sh | bash
Self-hosted
No data sent to us
AGPL-3.0 License