MCP Integration

Connect Converra to your AI coding assistant in 30 seconds. Use native OAuth where your client supports it; API keys remain available for intentional machine-to-machine integrations.

Quick Install

Choose your platform:

Claude Code

claude mcp add --transport http converra https://converra.ai/api/mcp
# Then authenticate from /mcp in Claude Code.

Codex

codex mcp add converra --url https://converra.ai/api/mcp
codex mcp login converra

Cursor

// Add to .cursor/mcp.json
{
  "mcpServers": {
    "converra": {
      "url": "https://converra.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

AntiGravity (Gemini)

// Add to ~/.gemini/antigravity/mcp_config.json
{
  "mcpServers": {
    "converra": {
      "url": "https://converra.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Claude Desktop

// Add to claude_desktop_config.json
{
  "mcpServers": {
    "converra": {
      "url": "https://converra.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Windsurf

{
  "mcpServers": {
    "converra": {
      "serverType": "sse",
      "url": "https://converra.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Native OAuth binds the connection to the tenant shown during consent. Use an API key only for an unattended or other machine-to-machine integration. Interactive access tokens expire after one hour. Compatible clients rotate refresh tokens within a 30-day grant; a revoked or expired grant requires a fresh browser login.

One-Click Install (Cursor)

Add to Cursor

After clicking, replace YOUR_API_KEY in Cursor settings with your actual key from Converra account settings.

What Can You Do?

After setup, just talk to your AI naturally:

Just Say	What Happens
"Here's my prompt, it's not working well"	Uploads to Converra and analyzes issues
"What's wrong with this prompt?"	Identifies weaknesses and improvement areas
"Test this against difficult users"	Runs optimization with simulated conversations
"How's the optimization going?"	Shows progress, variants being tested, early results
"What changed in the winning version?"	Explains the differences and improvements
"Apply it"	Deploys the optimized version
"Show me how my prompts are performing"	Displays metrics across all your prompts

Quick Example

Illustrative interaction — bringing your existing prompt to Converra:

You: Here's the prompt I use for my support bot. It's not working well -
     customers keep getting confused.
     [pastes existing prompt]

AI: I've uploaded your prompt to Converra. Looking at it, I see a few issues -
    it's too vague about when to escalate, and doesn't handle frustrated users well.

    Want me to run an optimization to find better versions?

You: Yes, test it against difficult customers

AI: Started an optimization. I'm simulating conversations with frustrated
    users, confused customers, and edge cases...

    Results are ready. Review the head-to-head comparison and decide whether
    to apply the candidate under your deployment policy.

You: What exactly changed?

AI: The candidate adds clarifying questions before responding. The comparison
    shows the evidence used to evaluate that change.

You: Apply it

AI: The candidate is ready to deploy according to your selected policy.

Try These Prompts

Copy any of these to get started:

"Here's my prompt - can you analyze what's wrong with it?" [paste your prompt]

"This prompt gets confused when users ask follow-up questions" [paste your prompt]

"Test my prompt against frustrated and impatient users"

"Here are some conversation logs - what's going wrong?" [paste your logs]

"Show me how my prompts are performing"

Next Steps

• Setup Guide - Detailed setup for all platforms
• Available Tools - All 25+ tools with examples