Model Context Protocol (MCP)

One Connection, Complete Access: Cekura’s MCP server provides both documentation search and API operations in a single endpoint. Ask questions about Cekura, search integration guides, create agents, run tests, and analyze results—all through natural language with your AI assistant.

Quickstart — OAuth in one click

For most users, MCP setup is two commands and a browser sign-in. No API keys, no env vars, no config files.

Claude Code
Claude Desktop

claude mcp add --transport http cekura --scope user https://api.cekura.ai/mcp

On first use, Claude Code opens a browser window — sign into your Cekura dashboard, click Authorize, and you’re done. The session token is managed by Claude Code automatically.Verify:

claude mcp list

Settings → Connectors → Add custom connector
Name: Cekura, URL: https://api.cekura.ai/mcp
Save → browser opens → sign in → done.

No config files, no Node.js, no keys.

That’s the entire setup. Skip to Verify Connection below. For Cursor, VS Code, Codex, or other clients, see the full configuration matrix below. For API-key-based auth (only needed for project-scoped credentials or shared CI access), expand the API Key path.

Prerequisites

Before setting up MCP, ensure you have:

Access to the Cekura Dashboard
Node.js 20.18.1 or higher installed — only if you’re configuring Cursor, VS Code, or another client that uses the mcp-remote shim. Claude Code and Claude Desktop don’t need Node.js.
- Check your version: node --version
- Download from: nodejs.org

Configure MCP Server (full matrix)

OAuth is the recommended path for nearly every user. Pick the API Key tab only if you need a project-scoped credential, want to share access via a key, or your client doesn’t support OAuth.

OAuth (recommended)
API Key

No header required. On first use, your client opens a browser window — sign into Cekura, click Authorize, and the session token is managed automatically.

Claude Code
Claude Desktop
Cursor / VS Code
Codex

claude mcp add --transport http cekura --scope user https://api.cekura.ai/mcp

Verify:

claude mcp list

Use the in-app Connectors UI — no config file editing required.

Open Claude Desktop → Settings → Connectors.
Click Add custom connector.
Set Name to Cekura and URL to https://api.cekura.ai/mcp.
Save. Claude Desktop opens a browser window to authorize against your Cekura dashboard account.
Once authorized, Cekura tools appear in the connector list and are available in any new chat.

No Node.js required for this path — Claude Desktop’s native connector handles the remote MCP transport directly.

Add to your Cursor MCP config or .vscode/mcp.json:

{
  "mcpServers": {
    "cekura": {
      "command": "npx",
      "args": ["mcp-remote", "https://api.cekura.ai/mcp"]
    }
  }
}

Add to ~/.codex/config.toml:

[mcp_servers.cekura]
url = "https://api.cekura.ai/mcp"

Use this if you want a project-scoped credential, want to share access via a key, or your client doesn’t support OAuth. Replace YOUR_API_KEY_HERE with the key from Settings → API Keys.

Claude Code
Claude Desktop
Cursor
VS Code
Codex

claude mcp add --transport http cekura --scope user https://api.cekura.ai/mcp --header "X-CEKURA-API-KEY:YOUR_API_KEY_HERE"

Verify:

claude mcp list

If you don’t specifically need an API key, use the OAuth tab above — Claude Desktop has a much simpler Connectors-UI flow that doesn’t require editing config files.

{
  "mcpServers": {
    "cekura": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://api.cekura.ai/mcp",
        "--header",
        "X-CEKURA-API-KEY:${CEKURA_API_KEY}"
      ],
      "env": {
        "CEKURA_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

{
  "mcpServers": {
    "cekura": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://api.cekura.ai/mcp",
        "--header",
        "X-CEKURA-API-KEY:${CEKURA_API_KEY}"
      ],
      "env": {
        "CEKURA_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Create or edit .vscode/mcp.json:

{
  "mcpServers": {
    "cekura": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://api.cekura.ai/mcp",
        "--header",
        "X-CEKURA-API-KEY:${CEKURA_API_KEY}"
      ],
      "env": {
        "CEKURA_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

CLI:

codex mcp add cekura --env CEKURA_API_KEY=YOUR_API_KEY_HERE -- sh -c 'npx -y mcp-remote https://api.cekura.ai/mcp --header "X-CEKURA-API-KEY:$CEKURA_API_KEY"'

Or, native HTTP via ~/.codex/config.toml:

[mcp_servers.cekura]
url = "https://api.cekura.ai/mcp"
http_headers = { "X-CEKURA-API-KEY" = "YOUR_API_KEY_HERE" }

Verify Connection

Restart your AI assistant and ask:

"List my Cekura agents"

If configured correctly, your assistant returns your agents from the Cekura API.

How MCP Works

Unified Access

One connection exposes 84+ tools spanning documentation search and API operations.

Documentation Search (no auth)

Search Cekura’s knowledge base for guides, schemas, and examples without an API key.

API Operations (auth required)

Create agents, run tests, analyze results, and manage your workspace via natural language.

Smart Tool Selection

Your assistant picks the right tool based on intent—answers questions or takes action.

Secure Authentication

OAuth uses a session-bound Bearer token your client manages automatically. API-key clients send X-CEKURA-API-KEY.

Working with MCP

Learn, then execute

Ask your assistant a question first, then take action with the same context:

"How do I integrate with Retell?" → searches docs
"Create a Retell agent named 'Support Bot' with assistant ID abc123" → creates it

Combine documentation with your data

“Show me how custom metrics work, then list the metrics on my Support Bot agent” — your assistant reads docs and your live workspace in one turn.

Iterate quickly with context

Your assistant maintains context across turns: list metrics → create a new one → generate test scenarios → run them.

For deeper end-to-end walkthroughs, see the Claude Code Guide.

Request Headers

Header	Used by	Description
`Authorization: Bearer <token>`	OAuth	Set automatically by your MCP client after sign-in
`X-CEKURA-API-KEY`	API key path	Required header when using an API key
`Content-Type: application/json`	Both	Request payload format

Tool Configuration

Tools are auto-generated from Cekura’s OpenAPI spec — every documented endpoint becomes an MCP tool with a typed inputSchema. You don’t author them; the server handles registration and stays in sync with the API. See the API Reference for endpoint details.

Best Practices

Auth hygiene

OAuth (preferred): Re-authorize when your session expires; revoke MCP access from the dashboard if a device is lost.
API key: Never commit keys to version control. Use environment variables. Rotate regularly. Use separate keys per environment. If you’re using an API key only because OAuth wasn’t obvious, switch — OAuth is simpler to revoke and avoids leaks via config files.

Handle rate limits

Cekura enforces per-plan rate limits. If you hit 429, slow down or upgrade your plan. Check usage in Settings → API Usage.

Test before production

Start with read-only operations (list agents, fetch results), then move to writes. Test in dev with a separate account or API key before pointing at production data.

Large responses (detailed transcripts, full call logs) consume context. Ask for filtered or summarized results when possible.

Available Tools

Cekura’s MCP server provides 84+ tools across four categories:

Documentation Search

No auth required

Search integration guides
Find API schemas and examples
Understand features and workflows

Testing

Auth required

Create and manage agents
Run test scenarios and evaluators
Define custom metrics
View and analyze results

Observability

Auth required

List and search call logs
Send calls for evaluation
Analyze call performance

Automation

Auth required

Create scheduled jobs
Manage cronjobs
Automate recurring tests

For per-endpoint details, see the API Reference.

Other MCP Clients

The Cekura MCP server speaks standard HTTP MCP, so any MCP-compatible client (Make.com, Zapier, custom servers, etc.) can connect to https://api.cekura.ai/mcp using the same OAuth or API-key flow.

Troubleshooting

Auth failed

Error: 401 Unauthorized

OAuth: Re-run the sign-in flow (your client usually re-prompts automatically). Confirm you’re signed into the right Cekura account in your browser.
API key: Verify the header name is X-CEKURA-API-KEY (case-sensitive) and the key is active in the dashboard. Generate a fresh key if needed.

Tools Not Showing Up

Issue: AI assistant doesn’t see Cekura tools

Restart your AI assistant after configuration
Verify the URL is https://api.cekura.ai/mcp
Check your config file syntax is valid JSON / TOML
Look for connection errors in the assistant’s logs

Node.js Version Error (Windows)

Error: Unsupported engine or ReferenceError: File is not definedCause: Multiple Node.js versions installed, wrong version being used.Solution:

List Node installs: where node
Find your Node v20+ path: node -e "console.log(process.execPath)"

Use the absolute path in your config:

{
  "mcpServers": {
    "cekura": {
      "command": "C:\\Full\\Path\\To\\node.exe",
      "args": [
        "C:\\Full\\Path\\To\\npx-cli.js",
        "-y",
        "mcp-remote",
        "https://api.cekura.ai/mcp"
      ]
    }
  }
}

Or uninstall older Node.js versions.

Rate Limit Exceeded

Error: 429 Too Many Requests

Check current usage in Settings → API Usage
Throttle requests in your workflow
Upgrade your plan or contact support for higher limits

References

API Reference

Complete API documentation

Cekura Skills

Install Cekura skills into your AI coding agent

Dashboard

Manage auth, projects, and API keys

Support

Get help with MCP setup

MCP

Guides

Documentation Index

​Quickstart — OAuth in one click

​Prerequisites

​Configure MCP Server (full matrix)

​Verify Connection

​How MCP Works

​Working with MCP

​Request Headers

​Tool Configuration

​Best Practices

​Available Tools

Documentation Search

Testing

Observability

Automation

​Other MCP Clients

​Troubleshooting

​References

API Reference

Cekura Skills

Dashboard

Support

Quickstart — OAuth in one click

Prerequisites

Configure MCP Server (full matrix)

Verify Connection

How MCP Works

Working with MCP

Request Headers

Tool Configuration

Best Practices

Available Tools

Other MCP Clients

Troubleshooting

References