Skip to content
Apploye

Connecting an AI client

Connect a supported AI client to Apploye over MCP. Read the MCP overview first: you need an active MCP-purpose key, and its granted scopes determine which tools the client can see.

Every client connects to this Streamable HTTP endpoint:

https://mcp.apploye.com/mcp

Set the key in your shell or operating-system secret environment before starting the client:

export APPLOYE_MCP_API_KEY="your MCP key"

Never commit the value, paste it into a project configuration file, or send it to support.

Claude Code

Create a user-scoped or project-scoped .mcp.json. Environment expansion keeps the key outside the file:

{
  "mcpServers": {
    "apploye": {
      "type": "http",
      "url": "https://mcp.apploye.com/mcp",
      "headers": {
        "X-APPLOYE-API-KEY": "${APPLOYE_MCP_API_KEY}"
      }
    }
  }
}

Start Claude Code from the shell where APPLOYE_MCP_API_KEY is defined. Use /mcp to inspect the connection and available tools.

Cursor

Add this to ~/.cursor/mcp.json for a private global configuration, or .cursor/mcp.json for a project configuration:

{
  "mcpServers": {
    "apploye": {
      "url": "https://mcp.apploye.com/mcp",
      "headers": {
        "X-APPLOYE-API-KEY": "${env:APPLOYE_MCP_API_KEY}"
      }
    }
  }
}

Open Cursor's MCP settings and confirm that apploye is connected and exposes only the tools allowed by the key's scopes.

Codex

Add the following to ~/.codex/config.toml, or to .codex/config.toml in a trusted project. The env_http_headers value is the environment-variable name, not the API key:

[mcp_servers.apploye]
url = "https://mcp.apploye.com/mcp"
env_http_headers = { "X-APPLOYE-API-KEY" = "APPLOYE_MCP_API_KEY" }
required = true

Restart Codex from the shell where APPLOYE_MCP_API_KEY is defined, then inspect its MCP server and tool list. Do not use codex mcp add --transport or --header; those flags are not the supported remote-header configuration contract.

OpenCode

Add a remote MCP entry to your OpenCode configuration. {env:APPLOYE_MCP_API_KEY} is resolved when OpenCode loads the file:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "apploye": {
      "type": "remote",
      "url": "https://mcp.apploye.com/mcp",
      "headers": {
        "X-APPLOYE-API-KEY": "{env:APPLOYE_MCP_API_KEY}"
      }
    }
  }
}

Compatibility matrix v1

This matrix is a documentation cutover gate, not a claim of unperformed verification. Before publishing developer.apploye.com, replace every pending cell with the exact tested client version, result, and UTC verification date. Each client must initialize, list the correctly scoped tools, and complete one real read operation.

ClientTested versionSetup mechanismResultVerified (UTC)
Claude CodePending cutover verification.mcp.json with environment-expanded headerPending: initialize, list tools, real readPending
CursorPending cutover verificationmcp.json with ${env:APPLOYE_MCP_API_KEY}Pending: initialize, list tools, real readPending
CodexPending cutover verificationconfig.toml with env_http_headersPending: initialize, list tools, real readPending
OpenCodePending cutover verificationRemote MCP config with {env:APPLOYE_MCP_API_KEY}Pending: initialize, list tools, real readPending

Verify the connection

For each client:

  1. Confirm the server initializes without an authentication error.
  2. List tools and verify that tools outside the MCP key's scopes are absent.
  3. Ask for one real, low-volume read, such as listing the organization's projects.
  4. Record the exact client version, result, and UTC date in the matrix above.

If a client fails, do not paste the key into a static header as a workaround. See Troubleshooting and contact Developer support with the request ID, client/version, and failure stage.

Related docs