CrewAI Agents Docker Image

TL;DR

1. Build image:

docker build --platform linux/amd64 --build-arg CREWAI=1.14.7 -t crewai:1.14.7 .

2. Start container with a project mount:

mkdir -p projects && \
docker run -it --rm --network host \
  -v "$PWD/projects:/home/appuser/apps" \
  -e P=my_project \
  crewai:1.14.7 \
  bash

3. Inside container, open your project:

v .
crewai run

Why this image

This image is a ready-to-use local environment for building and running CrewAI agents without local dependency setup.

It includes:

• neovim + LazyVim
• uv and crewai / crewai-tools
• fd, rg, fzf, and lazygit
• shell helpers (v, lg, sv, newcrew)

Prerequisites

• Docker
• Docker Compose (optional, for repeatable local deployment)
• Basic Python/CrewAI familiarity
• Ollama locally or remote API credentials (optional)

Quick start

Build the image

docker build \
  --platform linux/amd64 \
  --build-arg CREWAI=1.14.7 \
  -t sageil/crewai:1.14.7 .

Run a container

mkdir -p projects
docker run -it --rm --network host \
  -v "$PWD/projects:/home/appuser/apps" \
  -e P=my_project \
  sageil/crewai:1.14.7 \
  bash

The image defaults to P=default_crew if not provided.

Run with Compose

docker compose up -d
docker compose exec crewai bash

You can override image and project name:

IMAGE=sageil/crewai:latest \
CREWAI_PROJECT=my_project \
docker compose up -d

Working inside the container

Projects are created with crewai create crew on first startup (based on P).

Useful commands:

• v . → open current project in neovim
• newcrew <project_name> → create a new CrewAI project in current directory
• sv → activate .venv in current project (auto-run when present)
• crewai run → run the current project

LLM provider setup

Ollama on host

1. Open project in neovim (v .).
2. Open LazyVim explorer (SPACE+e) and show hidden files (SHIFT+H).
3. Update model/base settings to use your host endpoint (example: http://host.docker.internal:11434).

Remote providers (OpenAI/others)

1. Open your config in neovim.
2. Set OPENAI_API_KEY and model values in .env or crew config.
3. Run crewai run.

Code-based LLM override

from crewai.llm import LLM

myllm = LLM(
    model='ollama/deepseek-r1:7b',
    base_url='http://localhost:11434',
    temperature=0.2,
)

Assign llm=myllm in Agent(...) where needed.

Deployment notes

• Persist project outputs with ./projects:/home/appuser/apps.
• Remove network_mode: host in docker-compose.yml only when using external APIs and no host networking is required.
• Restart a stopped container with docker container start -ai <container_name>.

Supported versions

Recent and commonly used tags:

• 1.15.16
• 1.15.15
• 1.15.14
• 1.15.13
• 1.15.12
• 1.15.11
• 1.15.10
• 1.15.9
• 1.15.8
• 1.15.7
• 1.15.6
• 1.15.5
• 1.15.4
• 1.15.3
• 1.15.2
• 1.15.1
• 1.15.0
• 1.14.7
• 1.14.6
• 1.14.3
• 1.14.2
• 1.12.2
• 1.11.0
• 1.10.0
• 1.9.0
• 1.8.1
• 1.8.0
• 1.7.2
• 1.7.1
• 1.7.0

Full version history is available on Docker Hub:

• https://hub.docker.com/r/sageil/crewai/tags

Known issues

• Clipboard integration in neovim may require X11/display setup.
• Icon fonts may render inconsistently in some terminals.

File tree example

my_project/
├── .gitignore
├── .env
├── knowledge/
├── pyproject.toml
├── README.md
└── src/
    └── my_project/
        ├── __init__.py
        ├── main.py
        ├── crew.py
        ├── tools/
        │   ├── custom_tool.py
        │   └── __init__.py
        └── config/
            ├── agents.yaml
            └── tasks.yaml

Issues and support

• Report issues: https://github.com/sageil/crewai-docker-image/issues

License

• MIT: https://github.com/sageil/crewai-docker-image/blob/main/LICENSE.md