Back to Documentation

IDE & Agent Platforms

Connect Convext to 15+ agentic coding platforms and AI assistants.

Connect Convext to your AI coding assistants to automatically enforce engineering rules and technology standards while you code.

Integration Mechanisms

MCP Server (OAuth)

Real-time access via HTTP with automatic OAuth authentication

Generated Dotfiles

Static instruction files synced via CLI

Platform Support Matrix

Platform Vendor Type MCP Support Convext Integration
Cursor Anysphere IDE ✅ Full MCP + Dotfile
Claude Desktop Anthropic App ✅ Full MCP + Dotfile
Windsurf Codeium IDE ✅ Full MCP + Dotfile
Qoder Qoder IDE ✅ Full MCP
Kiro AWS IDE + CLI ✅ Full MCP
Replit Replit Cloud IDE ✅ Full MCP
Devin Cognition Agent ✅ Marketplace MCP
GitHub Copilot GitHub/MS IDE Plugin ✅ Full MCP + Dotfile
JetBrains AI JetBrains IDE Plugin ✅ Full MCP Server
Gemini CLI Google CLI ✅ Full MCP
Google Antigravity Google IDE ✅ Full MCP
Sourcegraph Cody Sourcegraph Extension ⚠️ Agentic MCP Context
Claude Code Anthropic CLI ✅ Full MCP + CLAUDE.md
Aider Aider CLI ✅ Full MCP + CLAUDE.md
✅ Full MCP

Complete MCP support with all Convext tools

⚠️ Partial

Limited MCP or special config required

❌ Dotfile Only

Use generated instruction files

MCP Server Integration

Direct HTTP Connection: Convext uses OAuth 2.1 for secure authentication. No npm packages or environment variables required!

What is MCP?

Model Context Protocol (MCP) is an open protocol that allows AI coding assistants to access external context providers dynamically.

Benefits

  • Real-time: Always up-to-date without manual syncing
  • Secure: OAuth 2.1 authentication with automatic token refresh
  • Zero Setup: No npm packages, no environment variables
  • Interactive: AI can update task status, report events

Available MCP Tools

Tool Description Use Case
get_rules Fetch engineering rules "Show me all critical security rules"
get_tech_standards Get technology standards "What's the Python testing standard?"
register_project Register new project "Register this repo"
get_project_context Full project context "What are the rules for this project?"

Claude Code Integration

Claude Code (by Anthropic) has native MCP support with OAuth 2.1 authentication.

Step 1: Configure MCP

Option A: Use the CLI command:

claude mcp add --transport http convext https://convext.app/mcp

Option B: Create a .mcp.json file in your project root:

{
  "mcpServers": {
    "convext": {
      "url": "https://convext.app/mcp",
      "transport": "http"
    }
  }
}

Step 2: Connect

On first connection, Claude Code will automatically:

  1. Discover OAuth endpoints via /.well-known/oauth-authorization-server
  2. Open your browser for authorization
  3. Log in with your Convext account (GitHub OAuth)
  4. Store and refresh tokens automatically

That's it! No API tokens to manage, no environment variables to configure. OAuth handles everything securely.

Step 3: Verify Connection

In Claude Code, type:

"Can you list the available Convext rules?"

Claude should use the get_rules tool to fetch rules from Convext.

Cursor Integration

Cursor has MCP support similar to Claude Code.

Step 1: Configure MCP

Create .mcp.json in your project root:

{
  "mcpServers": {
    "convext": {
      "url": "https://convext.app/mcp",
      "transport": "http"
    }
  }
}

Step 2: Generate .cursorrules

Cursor also reads .cursorrules for static context:

convext sync --format cursor

Best Approach: Use both MCP for dynamic queries and .cursorrules for quick reference and offline access.

GitHub Copilot Integration

GitHub Copilot reads .github/copilot-instructions.md automatically.

Step 1: Generate Instructions File

convext sync --format copilot

This creates .github/copilot-instructions.md with all your rules formatted for Copilot.

Step 2: Commit and Push

git add .github/copilot-instructions.md
git commit -m "Add Convext rules"
git push

Copilot automatically reads this file in your repository.

Keeping Instructions Up to Date

Re-sync when rules change:

# Check if out of sync
convext check

# Re-sync
convext sync --format copilot

# Commit changes
git add .github/copilot-instructions.md
git commit -m "Update Convext rules"

Windsurf

✅ Full MCP Support

Windsurf (formerly Codeium IDE) is an "agentic code editor" whose Cascade system watches your work and applies multi-file edits. Cascade supports MCP servers over HTTP.

Setup

Open Windsurf Settings → Cascade → MCP Servers:

{
  "mcpServers": {
    "convext": {
      "url": "https://convext.app/mcp",
      "transport": "http"
    }
  }
}

More MCP-Enabled Platforms

All platforms below support the same HTTP transport configuration:

{
  "mcpServers": {
    "convext": {
      "url": "https://convext.app/mcp",
      "transport": "http"
    }
  }
}
Gemini CLI ✅ Full MCP

Google's open-source AI agent for your terminal. Supports MCP servers for extended functionality. See Gemini CLI docs.

Google Antigravity ✅ Full MCP

Google's AI-powered IDE with agent-first architecture. Agents autonomously control editor, terminal, and browser. See antigravity.codes.

JetBrains AI Assistant ✅ Full MCP

JetBrains IDEs now have integrated MCP servers, allowing external clients to access IDE tools. See JetBrains MCP docs.

Qoder ✅ Full MCP

Agentic coding platform with MCP Square for browsing servers. Supports HTTP and SSE transports.

Kiro (AWS) ✅ Full MCP

AWS's agentic IDE for "spec-driven development". MCP support is a core feature for connecting specialized tools.

Replit ✅ Full MCP

Cloud IDE that adopted MCP as a "universal connector" for AI tools. See their MCP docs.

Devin ✅ MCP Marketplace

AI software engineer with MCP Marketplace. Connects to external tools via MCP using HTTP transports.

Sourcegraph Cody ⚠️ Agentic Context

"Agentic chat" that autonomously pulls context from codebase, terminal, and MCP tools before answering.

CLI Tools

✅ Full MCP Support: Both Claude Code and Aider support MCP servers natively. Use MCP for real-time rules, CLAUDE.md as fallback.

Claude Code

✅ MCP

Anthropic's agentic CLI. Create .mcp.json:

{
  "mcpServers": {
    "convext": {
      "url": "https://convext.app/mcp",
      "transport": "http"
    }
  }
}

Aider

✅ MCP

CLI coding assistant with MCP support. Configure in .aider.conf.yml:

mcp-servers:
  - name: convext
    url: https://convext.app/mcp
    transport: http

Best Practices

1. Keep Rules Synced

Problem: Stale dotfiles don't reflect updated rules.

Solution: Sync regularly

# Daily (in bin/dev):
convext sync

# Or add to package.json:
npm run charter:sync

2. Commit Dotfiles (Team Decision)

Pros of Committing

  • • Team shares same rules
  • • No setup required for new devs
  • • Visible in code review

Cons of Committing

  • • Files can get out of sync
  • • Merge conflicts on rule updates
  • • Repo bloat

Troubleshooting

OAuth Authorization Not Opening

Symptoms: Browser doesn't open for authorization

Solutions:

  • Ensure your browser allows popups from the MCP client
  • Try manually visiting https://convext.app/oauth/authorize
  • Check that https://convext.app is accessible

MCP Connection Failed

Symptoms: "MCP server unavailable" error in IDE

Solutions:

  • Verify .mcp.json uses "transport": "http"
  • Check that https://convext.app is accessible
  • For Claude Code: Clear stored MCP credentials and reconnect
  • Restart your IDE

AI Not Following Rules

Symptoms: Suggestions ignore your standards

Solutions:

  • Run convext check to verify sync status
  • Check dotfile exists in expected location
  • Re-sync: convext sync --force
  • Explicitly mention rules in prompts

Outdated Rules

Symptoms: Old rules still appearing

Solutions:

  • Sync with --force flag to overwrite
  • Remove dotfiles and re-sync
  • Check if you're using MCP (always current) vs dotfiles
Core Concepts Admin Guide