Open Source

Build AI Agent
Workflows in YAML

Rapidly prototype multi-model AI workflows with file analysis, multi-agent interactions, and Claude Code integration. Everything version-controlled, transparent, and experimental.

1.2k stars
89 forks
2.1k commits

Get Exclusive Elf0 Resources

Join our community for curated workflows, marketplace updates, and expert guidance:

Curated YAML agent templates and workflow examples
First updates about our upcoming AI agent marketplace
Expert tutorials and workflow optimization guides
Community insights and development updates

Join the Elf0 Community

Get curated resources and marketplace updates

    No spam, ever. We respect your privacy and you can unsubscribe at any time.
    Built with Kit

    Why Developers Choose Elf0

    Built for rapid prototyping and production-ready AI agent workflows

    Multi-Model Freedom

    Switch between OpenAI, Anthropic, and Ollama models in the same workflow.

    Compare outputs, combine strengths, avoid vendor lock-in. True model flexibility.

    File-Aware Agents

    Reference entire codebases, documents, or directories with simple @ syntax.

    Your agents understand your project context automatically.

    Multi-Agent Orchestration

    Build workflows where multiple AI agents collaborate, debate, and refine each other's work.

    Create sophisticated agent interactions and simulations.

    Claude Code Integration

    Seamlessly connect with Claude Code and MCP (Model Context Protocol).

    Build agents that integrate with your development environment.

    Experimental & Transparent

    See exactly what your agents are doing.

    Full transparency in prompts, model calls, and reasoning. Perfect for research and experimentation.

    YAML Simplicity

    No complex frameworks or APIs.

    Everything in readable YAML that you can version, share, and improve with AI assistance. Simple but powerful.

    From AI Framework Complexity to Simple YAML

    Ever spent days setting up LangChain or custom API integrations just to test an AI agent idea? Elf0 lets you experiment with multi-model workflows, file analysis, and agent interactions using simple YAML configuration. No vendor lock-in, full transparency.

    Multiple AI models (OpenAI, Anthropic, Ollama)
    File analysis with @ syntax
    Multi-agent workflows and simulations
    Claude Code and MCP integration
    simple-chat.yaml YAML
    version: "v1"
description: "Basic chat workflow"
runtime: "langgraph"

llms:
  chat_llm:
    type: "openai"
    model_name: "gpt-4.1-mini"
    temperature: 0.7
    params:
      max_tokens: 1000
      system_prompt: |
        You are a friendly assistant who always ends a conversation with an emoji.

workflow:
  type: "sequential"
  nodes:
    - id: "chat_step"
      kind: "agent"
      ref: "chat_llm"
      stop: true
  edges: []

    Real-World Example

    See how Elf0 handles complex document processing workflows

    Multi-Agent Code Review System

    Build a comprehensive code review system using multiple AI models that analyze security, performance, and maintainability across your entire codebase.

    1

    Codebase Analysis

    Claude analyzes entire project structure with @src/ syntax

    2

    Security Audit

    GPT-4 performs deep security vulnerability assessment

    3

    Agent Collaboration

    Multiple agents debate findings and create consensus report

    Terminal Output Running
    $ elf0 run multi-agent-reviewer.yaml
    > Loading workflow: Multi-Agent Code Review v2.1
    > Claude-4: Analyzing project structure @src/ (2.3s)
    > GPT-4o: Security scan on 47 files (5.1s)
    > Claude-4: Cross-referencing findings (1.8s)
    > GPT-4o: Generating prioritized recommendations (2.4s)
    > ✓ Multi-agent review completed
    > Report: comprehensive_review.md (3,247 words)

    Get Started in Minutes

    1. Clone and install (5 mins)

    git clone https://github.com/emson/elf0.git && cd elf0

    2. Set up Python environment

    uv venv && source .venv/bin/activate && uv pip install -e .

    3. Generate a question-finder agent

    uv run elf0 -v agent specs/utils/creator_workflow_v2.yaml \
  --prompt "Create a new YAML agent that when given a prompt \
  will analyse it, identify key concepts, then reason hard \
  about what questions would could be asked about them." \
  --output questions.yaml

    Elf0 Frequently Asked Questions