Crew Builder

Build a team of AI agents in a web UI, then download a project you can run.

No need to hand-write CrewAI configs. Click around, hit Build, unzip, run.

Why this project

Crew Builder produces a multi-agent system with no external dependencies — no databases, no embedding stores, no vector DBs, no shared infrastructure you have to stand up and operate.

What you download is a self-contained project: agents, skills, knowledge, and tools packaged to run on their own (expect external MCPs). That keeps the stack simple and lets you scale in many ways — more agents in a crew, more crews side by side, more machines or containers running the same zip — without tying growth to a central database or embedding service. You can use the downloaded project as a good boilerplate for your work.

Run it

You need Docker.

git clone https://github.com/abdollahpour/crewbuilder.git
cd crewbuilder
docker compose up --build

Open http://localhost:8080.

There's a sample travel-assistant crew already loaded so you can poke around.

Make a crew and run it

First of all, you need Ollama. Install and run:

ollama run qwen3:8b

The code shipped by some sample data-set.

1. Install CrewAI
2. Create agents (who does what).
3. Optionally add skills, knowledge, or MCPs (extra tools / docs).
4. Create a crew and add your agents.
5. Click Build → download the zip.
6. Unzip it
7. (Optional if you used cloud provider, default is ollama) copy .env.example to .env, put in your API keys.
8. Run:

crewai run

Try to feed some samples. Ex:
What’s the weather like in Berlin? or I’m planning to travel from London to Berlin in summer. I’d like to time the trip around a good music festival. Please find the best dates, along with suitable flights and a hotel, keeping my total budget around €2,000.

Run it. It takes time, specially if you are using your local machine and Ollama. Your answer lands in output/report.md.

Words we use

• Agent — one AI worker with a job (CrewAI agents)
• Crew — a team of agents plus a boss that delegates (CrewAI crews)
• Skill — reusable instructions an agent can follow (Agent Skills)
• Knowledge — docs the agent can search (packaged as a local FAISS index in the download)
• MCP — an external tool server the agent can call (Model Context Protocol)

References

• CrewAI — multi-agent framework the downloaded project runs on
• Agent Skills — open format for portable agent instructions (SKILL.md)
• Model Context Protocol (MCP) — standard for connecting agents to external tools
• Ollama — local LLM runtime (default provider; sample uses qwen3:8b)
• uv — Python package runner used to launch the generated crew
• Docker — runs the Crew Builder stack itself

License

MIT © 2026 Hamed Abdollahpour