repository·master·Indexed 27 days ago
https://github.com/kyegomez/swarmsAn enterprise-grade multi-agent orchestration framework for reliable, scalable, and adaptive agentic workflows. Swarms supports sequential, concurrent, and hierarchical architectures, integrating with the Model Context Protocol (MCP) and various LLMs. It provides a comprehensive CLI for managing agents, running complex workflows like LLM Council and HeavySwarm, and automating configurations via YAML or Markdown.
To run the provided shell script examples in the examples/cli/ directory, make them executable and run the setup check script:
chmod +x *.sh
./01_setup_check.shAlternatively, you can execute them directly using bash:
bash 01_setup_check.shTo help AI coding assistants (Claude Code, Cursor, Windsurf, Continue, etc.) write idiomatic Swarms code without extra prompting, you can use the CLAUDE.md file provided in the repository.
Action: Drop CLAUDE.md (or symlink it as AGENTS.md or .cursorrules) into your project directory. This file contains instructions on the Agent primitive, multi-agent architectures (including SequentialWorkflow, ConcurrentWorkflow, AgentRearrange, GraphWorkflow, MixtureOfAgents, HierarchicalSwarm, and SwarmRouter), tools, streaming, memory, and MCP integration.
Before pushing to GitHub to trigger the automated workflow, you can verify your Docker configuration by building and running the image locally.
# Build the image locally
docker build -t swarms:test .
# Test the image
docker run --rm swarms:test python test_docker.pySwarms provides multiple documentation formats depending on your needs:
llms.txt): A single, machine-readable index designed for AI coding assistants (like Cursor, Claude Code, or Windsurf) to consume the entire documentation in one fetch. Point your AI tool to https://docs.swarms.world/llms.txt to enable idiomatic code generation.Agent, SequentialWorkflow, ConcurrentWorkflow, AgentRearrange, GraphWorkflow, and SwarmRouter at docs.swarms.world/api.To enable the automatic building and pushing of Docker images via GitHub Actions, you must configure two specific repository secrets in GitHub. This requires a DockerHub account and an access token with 'Read & Write' permissions.
| Secret Name | Value | Description |
|---|---|---|
DOCKERHUB_USERNAME | Your DockerHub username | Your DockerHub username (e.g., kyegomez) |
DOCKERHUB_TOKEN | Your DockerHub access token | The access token created in DockerHub Security settings |
To install from the source code, clone the repository and install the requirements manually.
# Clone the repository
$ git clone https://github.com/kyegomez/swarms.git
$ cd swarms
$ pip install -r requirements.txtManage your Swarms configuration and credentials using the following commands:
swarms onboarding starts the interactive onboarding process.swarms get-api-key retrieves your configured API keys.swarms check-login verifies your current login status.swarms onboarding
swarms get-api-key
swarms check-loginFor production stability, pin your deployment to a specific version tag instead of using latest.
# Pull a specific version
docker pull kyegomez/swarms:v8.0.4
# Run with specific version
docker run --rm kyegomez/swarms:v8.0.4 python -c "import swarms; print(swarms.__version__)"To build the image from source, ensure you have Docker and Git installed. Clone the repository and use the docker build command.
# Clone the repository
git clone https://github.com/kyegomez/swarms.git
cd swarms
# Build the image
docker build -t swarms:latest .
# Test the image
docker run --rm swarms:latest python test_docker.pySwarms supports the Model Context Protocol (MCP), allowing you to connect a Swarms Agent to any MCP server. This enables the agent to automatically discover and call tools provided by the server.
For detailed implementation steps, refer to the MCP Integration Guide.
Setting max_loops="auto" allows an agent to decide when a task is complete. The agent will continue reasoning and acting until it reaches a stopping condition. This is ideal for open-ended research or iterative tasks (e.g., write $\rightarrow$ review $\rightarrow$ revise).
For latency or cost-sensitive production pipelines with well-defined steps, use a fixed integer for max_loops instead.
from swarms import Agent
agent = Agent(
agent_name="Autonomous-Research-Agent",
agent_description="An autonomous agent that conducts multi-step research independently.",
system_prompt=(
"You are an autonomous research agent. Break down complex tasks into steps, "
"execute each step thoroughly, and signal completion only when the full task is done."
),
model_name="gpt-5.4",
max_loops="auto", # Agent decides when it's done — no fixed iteration cap
autosave=True,
verbose=True,
)
# The agent will keep looping — planning, executing, and reflecting — until it
# determines the task is fully complete.
result = agent.run(
"Research the current state of quantum computing, identify the top three "
"hardware approaches, and summarize the key challenges each faces."
)
print(result)You can quickly pull and run the latest Swarms image from DockerHub. Use the latest tag for general testing or run an interactive shell to explore the environment.
# Pull the latest image
docker pull kyegomez/swarms:latest
# Run a simple test
docker run --rm kyegomez/swarms:latest python test_docker.py
# Run with interactive shell
docker run -it --rm kyegomez/swarms:latest bash