automagent import

Convert agent definitions from CrewAI, OpenAI, and other frameworks

Usage

$ automagent import <path> [options]

The <path> argument is required and points to the source file you want to convert.

Options

Format auto-detection

When you don’t specify --format, the CLI inspects the source file to determine its format:

If auto-detection fails, use --format to specify explicitly.

Field mapping

CrewAI

OpenAI Agents SDK

LangChain

Fields that don’t map cleanly are preserved under the extensions namespace so nothing is lost. Auto-generated descriptions and any fields that need human review are marked with # TODO: Review comments in the output.

Examples

Import a CrewAI agent:

$ automagent import ./crewai-agent.yaml

Given a CrewAI file like:

role: Senior Data Analyst
goal: Analyze sales data and identify trends
backstory: >
  You are an experienced data analyst with 10 years
  of experience in retail analytics.
llm: gpt-4
tools:
  - search_tool
  - csv_reader

The resulting agent.yaml:

apiVersion: v1
kind: agent

name: senior-data-analyst
version: 0.1.0
description: Senior Data Analyst # TODO: Review

model: gpt-4

instructions: |
  Goal: Analyze sales data and identify trends

  Backstory: You are an experienced data analyst with 10 years
  of experience in retail analytics.

tools:
  - name: search_tool
  - name: csv_reader

extensions:
  crewai:
    role: Senior Data Analyst

Import an OpenAI assistant:

$ automagent import ./assistant.json -o research-agent.yaml

Given an OpenAI assistant JSON:

{
  "name": "research-assistant",
  "instructions": "You help researchers find and summarize academic papers.",
  "model": "gpt-4o",
  "tools": [
    { "type": "code_interpreter" },
    { "type": "file_search" }
  ]
}

The resulting agent.yaml:

apiVersion: v1
kind: agent

name: research-assistant
version: 0.1.0
description: research-assistant # TODO: Review

model: gpt-4o

instructions: |
  You help researchers find and summarize academic papers.

tools:
  - name: code_interpreter
  - name: file_search

extensions:
  openai:
    tools:
      - type: code_interpreter
      - type: file_search

Import a LangChain agent:

$ automagent import ./langchain-agent.json

Given a LangChain config like:

{
  "name": "Research Agent",
  "system_message": "You help researchers find and summarize papers.",
  "llm": {
    "model": "gpt-4",
    "temperature": 0.7
  },
  "tools": [
    { "name": "search_papers", "description": "Search academic papers" }
  ],
  "tags": ["research", "academic"]
}

The resulting agent.yaml:

apiVersion: v1
kind: agent

name: research-agent
version: 0.1.0
description: Imported from LangChain agent: Research Agent # TODO: Review

model:
  id: gpt-4
  settings:
    temperature: 0.7

instructions: |
  You help researchers find and summarize papers.

tools:
  - name: search_papers
    description: Search academic papers

metadata:
  tags:
    - research
    - academic

Force a specific format:

$ automagent import ./config.json --format openai --force

Custom output path:

$ automagent import ./legacy-agent.yaml -o agents/converted.yaml

Notes

• Schema validation runs automatically after import. You will see any warnings or errors in the generated file immediately.
• Review # TODO: Review comments in the output. These mark auto-generated fields (like description) and unmapped values that may need adjustment.
• The name field is automatically slugified from the source (lowercased, spaces replaced with hyphens, invalid characters removed).
• The command will not overwrite an existing output file unless --force is passed.
• Unmapped fields are preserved in extensions.<framework> so you can reference the original values.

See also

• Import guide — Step-by-step walkthrough with more detail
• automagent validate — Review warnings after import
• Field reference — Understand the target format