We built AI Integration to make complex broadcasting and panel workflows faster and easier to discover. Instead of moving through several screens or learning API calls, operators can investigate audience behaviour, organize media, update playlists, monitor stations, and perform administrative work through a conversation. The quality of the result still depends on the AI model, your prompt, and the permissions you grant it.
MediaCP AI Integration connects your panel to AI assistants through a Model Context Protocol (MCP) server that ships with MediaCP for Unlimited subscriptions. It is compatible with Claude, Codex, the ChatGPT desktop app, LM Studio, Windsurf, and other clients that support remote MCP servers over Streamable HTTP.
System Requirements
- An active MediaCP Unlimited subscription or Perpetual license.
- MediaCP 2.16 or newer
- API Access and MCP Server enabled under System Config > Misc.
- The MediaCP Port 80/443 Proxy service installed and enabled.
- A panel URL that uses HTTPS and does not contain an explicit port, for example
https://panel.example.com/rather thanhttps://panel.example.com:2020/. - A publicly reachable panel URL for cloud-hosted AI clients such as Claude.
- An MCP-compatible AI client and a model with reliable tool-calling support.
Examples
The integration exposes tools for reporting, services, media, playlists, schedules, listeners and viewers, stream targets, users, announcements, DJs, jingles, events, and system statistics.
- When do listeners spend the most time on My Hits Radio? Compare the last four weeks by day and hour.
- Find the most popular tracks on My Hits Radio this month and propose a 25-track playlist. Show me the proposed tracks before creating it.
- Show me what is playing now, the current listener count, and the status of every stream target for service 105.
- Read the last 20 tracks played by service 105 and look for repetition or gaps that could explain the station complaint.
- Draft a maintenance announcement for tomorrow at 10:00 AM. Show me the wording and audience settings before publishing it.
Connecting to your AI client
- Sign in to MediaCP.
- If you are an administrator, open System Config > Misc and confirm that API Access and MCP Server are enabled.
- Open Integrations from the side menu.
- Enable API access for your account if prompted.
- Copy the MCP Server URL and your Integration Key. The URL normally has this format:
https://panel.example.com/mcp
Connect Claude
Claude custom connectors support remote MCP servers. See Claude’s custom connector guide for current plan and workspace controls.
In an individual Claude account, open Customize > Connectors. For a Team or Enterprise workspace, an Owner must first open Organization settings > Connectors.
Select Add custom connector. In an organization workspace, select Custom, then Web.
Enter a name such as
MediaCPand paste the MCP Server URL from the MediaCP Integrations page.Open Advanced settings and enter:
Setting Value OAuth Client ID mediacpOAuth Client Secret Your MediaCP Integration Key Save the connector, select Connect, and complete the authorization flow.
In a conversation, open the + menu, choose Connectors, and enable MediaCP.
If Claude asks for explicit OAuth endpoints, use:
Authorization URL: https://panel.example.com/mcp/authorize
Token URL: https://panel.example.com/mcp/token
Claude connects from its cloud infrastructure, including when you use Claude Desktop. Your MediaCP URL must therefore be publicly reachable over HTTPS.
Connect Codex or ChatGPT desktop
Codex and the ChatGPT desktop app share MCP configuration and support Streamable HTTP with bearer-token authentication. See the official MCP configuration guide.
Option 1: Codex CLI
Set the Integration Key in an environment variable, then add MediaCP:
export MEDIACP_INTEGRATION_KEY="YOUR_INTEGRATION_KEY"
codex mcp add mediacp \
--url "https://panel.example.com/mcp" \
--bearer-token-env-var MEDIACP_INTEGRATION_KEY
codex mcp get mediacp
Make MEDIACP_INTEGRATION_KEY available to future Codex sessions through your operating system’s secure environment configuration. Do not commit the key to a repository.
Option 2: Codex configuration file
Add this to ~/.codex/config.toml:
[mcp_servers.mediacp]
url = "https://panel.example.com/mcp"
bearer_token_env_var = "MEDIACP_INTEGRATION_KEY"
Restart Codex or the ChatGPT desktop app after saving the configuration. In the ChatGPT desktop app, you can also open Settings > MCP servers, add a Streamable HTTP server, and enter the same URL. Use the shared configuration file when bearer-token settings are not shown in the interface.
ChatGPT web
ChatGPT web custom apps currently require an MCP-compliant OAuth 2.1 authorization flow with discovery metadata and PKCE. ChatGPT does not accept a custom static API key for this connection. The current MediaCP server provides a simpler client ID and secret exchange for Claude and does not yet implement all of ChatGPT web’s OAuth 2.1 requirements. Use Codex or the ChatGPT desktop app for the current MediaCP release. See OpenAI’s MCP authentication requirements for the web integration requirements.
Connect LM Studio
LM Studio can connect local models to remote MCP servers. Use a current LM Studio release and choose a model that is good at tool calling. See the LM Studio MCP guide for the latest interface.
Open the Program tab in the right sidebar.
Select Install > Edit mcp.json.
Add the MediaCP server:
{ "mcpServers": { "mediacp": { "url": "https://panel.example.com/mcp", "headers": { "Authorization": "Bearer YOUR_INTEGRATION_KEY" } } } }Save the file. LM Studio should load the server and show its tools.
Start a chat with a tool-capable model, enable the MediaCP tools, and approve calls when prompted.
This configuration stores the Integration Key in mcp.json. Protect the file with your operating system account permissions and never share or commit it.
Troubleshooting
The AI gives incomplete or poor answers
We have documented every MCP-exposed part of MediaCP internally so the AI receives detailed tool descriptions, parameters, natural-language usage guidance, and examples. There will always be room for improvement, and the integration is only as capable as the AI using it. Choose a model that performs well at everyday reasoning and tool-calling tasks, give it a clear goal, name the station or service, and include the time period you want analyzed.
Reporting requests are deliberately rate limited and a single reporting query covers no more than 31 days. Ask the AI to split longer comparisons into several periods instead of repeatedly retrying one large request.
The AI client cannot connect
Confirm that the client is using the exact MCP Server URL shown on the MediaCP Integrations page. The URL must end in
/mcpand must not include the panel’s old port number.Confirm that the header is
Authorization: Bearer YOUR_INTEGRATION_KEY, with one space afterBearer.Confirm that API Access is enabled for both the system and the MediaCP user, and that MCP Server is enabled under System Config > Misc.
Sign in as the root user over SSH and check all MediaCP services:
mediacp statusYou can also check the MCP Server directly:
mediacp status mcpserverFrom the MediaCP server, check the local health endpoint:
curl http://127.0.0.1:3000/healthA running server returns JSON with
"status":"ok".
The MCP Server will not start, crashes, or has another problem
Logs are stored in:
/usr/local/mediacp/log/mcpserver/
Check the latest mcp-server-YYYY-MM-DD.log, mcp-error-YYYY-MM-DD.log, and mcp-server.out files. To restart the service in foreground debug mode and display detailed requests and errors, run as root:
mediacp restart mcpserver debug
Press Ctrl+C after collecting the error, then start the service normally:
mediacp start mcpserver
Never publish debug output without reviewing it first because requests and error details may contain private station data.
Tools are missing or an action is denied
MediaCP filters tools according to the user level associated with the Integration Key. Customer keys receive customer-level tools, while reseller and administrator tools require the corresponding account permissions. Sign in as the intended MediaCP user, copy that user’s key from Integrations, and reconnect the AI client.