Convert

Convert Claude Code Workflows to Other AI Tools

Convert Claude Code artifacts — CLAUDE.md, hooks, Anthropic SDK code — to equivalent configs for Gemini CLI, GitHub Copilot, Cursor, or OpenAI Codex.

When to Apply

Trigger when the user:

• Says "convert to Gemini", "migrate to Copilot", "port to Cursor", "translate to Codex/OpenAI"
• Wants to replicate their Claude Code setup in another AI coding tool
• Has Anthropic SDK code they want to migrate to another provider

Workflow

Step 1: Detect Claude Artifacts

Scan the project for Claude-specific artifacts:

# Config files
ls CLAUDE.md .claude/CLAUDE.md .claude/settings.json .claude/settings.local.json 2>/dev/null

# Hooks
ls .claude/hooks/ 2>/dev/null

# Anthropic SDK usage
grep -r "from anthropic\|import anthropic\|@anthropic-ai/sdk\|Anthropic()" \
  --include="*.py" --include="*.ts" --include="*.js" . 2>/dev/null | head -20

Report what was found before proceeding.

Step 2: Identify Target Tool

If not specified, ask:

Which tool would you like to convert to?

1. Google Gemini CLI
2. GitHub Copilot
3. Cursor
4. OpenAI Codex

Step 3: Fetch Current Docs via context7

REQUIRED — never skip, never use training-data guesses.

Fetch current documentation using context7 before generating any config or SDK translation:

Gemini CLI:

mcp__plugin_context7_context7__resolve-library-id: "gemini cli"
mcp__plugin_context7_context7__query-docs: config format, GEMINI.md, system prompts, settings.json

GitHub Copilot:

mcp__plugin_context7_context7__resolve-library-id: "github copilot"
mcp__plugin_context7_context7__query-docs: copilot-instructions.md format, custom instructions

Cursor:

mcp__plugin_context7_context7__resolve-library-id: "cursor"
mcp__plugin_context7_context7__query-docs: .cursorrules format, cursor rules, .cursor/rules

OpenAI Codex / AGENTS.md:

mcp__plugin_context7_context7__resolve-library-id: "openai"
mcp__plugin_context7_context7__query-docs: AGENTS.md format, openai SDK chat completions

Cite the resolved library ID and doc version in every generated file header.

Step 4: Generate Target Configs

CLAUDE.md → Target Config

Read CLAUDE.md fully (expanding any @file imports). Then write:

Claude artifact	Gemini CLI	GitHub Copilot	Cursor	OpenAI Codex
CLAUDE.md	GEMINI.md	.github/copilot-instructions.md	.cursorrules	AGENTS.md
.claude/settings.json hooks	.gemini/settings.json	⚠️ No equivalent	.cursor/rules/	⚠️ No equivalent

Add a header comment to every generated file:

# Generated by /claude4ops:convert
# Source: CLAUDE.md
# Target: [Tool Name]
# Docs: context7 [library-id] — fetched [date]
# Review before committing. Claude-specific sections marked ⚠️.

Keep the original CLAUDE.md intact — generate new files alongside, never overwrite.

Anthropic SDK → Target SDK

For each file with Anthropic SDK imports, output a side-by-side migration:

Python — Anthropic → Google Generative AI (Gemini):

# BEFORE (Anthropic)
import anthropic
client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-opus-4-8",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}]
)
print(message.content[0].text)

# AFTER (google-generativeai)
import google.generativeai as genai
import os
genai.configure(api_key=os.environ["GEMINI_API_KEY"])
model = genai.GenerativeModel("gemini-2.0-flash")  # adjust model as needed
response = model.generate_content("Hello")
print(response.text)

Python — Anthropic → OpenAI (Copilot / Codex):

# BEFORE (Anthropic)
import anthropic
client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-opus-4-8",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}]
)
print(message.content[0].text)

# AFTER (openai)
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",  # adjust model as needed
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

TypeScript — Anthropic → OpenAI:

// BEFORE
import Anthropic from "@anthropic-ai/sdk";
const client = new Anthropic();
const msg = await client.messages.create({
  model: "claude-opus-4-8",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello" }],
});
console.log(msg.content[0].text);

// AFTER
import OpenAI from "openai";
const client = new OpenAI();
const response = await client.chat.completions.create({
  model: "gpt-4o",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello" }],
});
console.log(response.choices[0].message.content);

Model name mapping (approximate):

Anthropic	Gemini	OpenAI
claude-opus-4-x (most capable)	gemini-2.0-pro	gpt-4o
claude-sonnet-4-x (balanced)	gemini-2.0-flash	gpt-4o-mini
claude-haiku-4-x (fastest)	gemini-2.0-flash-lite	gpt-4o-mini

Always verify model availability against current docs fetched from context7.

Hook Mappings

Claude Code hook	Gemini CLI	Copilot	Cursor	Codex
PreToolUse (block commands)	.gemini/settings.json extensions	⚠️ No equivalent	Cursor Rules	⚠️ No equivalent
Stop (auto-lint)	⚠️ No equivalent	⚠️ No equivalent	⚠️ No equivalent	⚠️ No equivalent
PostToolUse (audit log)	⚠️ No equivalent	⚠️ No equivalent	⚠️ No equivalent	⚠️ No equivalent
Notification hooks (Slack)	⚠️ No equivalent	⚠️ No equivalent	⚠️ No equivalent	⚠️ No equivalent

Step 5: Flag Unsupported Features

For every Claude feature with no equivalent, include a warning block in the generated output:

⚠️  UNSUPPORTED: Claude Code hooks (PreToolUse / Stop / PostToolUse)
    These run shell scripts at tool-call boundaries. [Target] has no equivalent.
    Workaround: Use git hooks (pre-commit) or CI pipeline steps instead.

⚠️  UNSUPPORTED: Tool-use permissions (allowedTools / blockedTools)
    Claude Code's fine-grained tool permission system has no direct equivalent.
    Workaround: Document rules in the target config file as natural-language instructions.

⚠️  UNSUPPORTED: MCP server configuration
    MCP is Claude Code-specific. Some tools support OpenAPI/function calling instead.
    Workaround: Expose your MCP server via OpenAPI spec and register as a tool/function.

⚠️  UNSUPPORTED: claude4ops plugin skills and commands
    These are Claude Code-only. No equivalent exists in other tools.

Step 6: Output Summary

After generating all files, print:

Conversion complete: Claude Code → [Target Tool]
Docs fetched: context7 [library-id] (fetched [date])

Generated:
  ✓ [target config file]       ← from CLAUDE.md
  ✓ [sdk-migration.md]         ← Anthropic SDK migration notes

Warnings ([N] features with no equivalent):
  ⚠️  hooks → no equivalent in [target tool]
  ⚠️  MCP servers → no equivalent (see workaround above)
  ⚠️  tool permissions → no equivalent

Review all generated files before committing.
Source files were not modified.

Notes

• Never generate target configs without first fetching context7 docs — formats change
• Expand @file imports in CLAUDE.md before mapping content to target format
• Source files (CLAUDE.md, SDK files) are read-only — generate new files alongside
• For multi-target conversions, run Steps 3–6 once per target tool
• If context7 cannot resolve a library, warn the user and halt rather than guessing