MCP

Model Context Protocol integration for AI agents that read from Conviva.

Updated 2026-08-04 mcp, model context protocol, agents, ai

Conviva Connect MCP (Model Context Protocol) exposes Conviva data to AI agents so Conviva customers can quickly and securely request Conviva data for MCP agent tasks. Whether providing a Support team with issue context before a Help Desk conversation commences with viewer operational data or automating operational performance reports, Conviva Connect MCP provides the foundation for extremely rich AI applications powered by Conviva data.

How Conviva MCP works: a customer agent connects tools like desktop LLMs, dashboards, and reports to Conviva data through the MCP server

Conviva MCP lets an AI agent bring Conviva data into your own tools, such as desktop LLMs, operational dashboards, service management, and content or go/no-go reports. Use it when you want to query Conviva data through natural language or build agents that act on it.

The Conviva MCP server enables your AI agents to analyze and interact with Conviva's Video Streaming Insights (VSI) Video and App. The Conviva MCP server provides a set of tools that AI agents use to query video metrics, investigate AI-powered alerts, explore viewer sessions, and analyze digital product experience data, all through natural language prompts and agent-to-agent communication.

Common MCP use cases include generating reports and dashboards that leverage the power of compiling Conviva VSI Video and App data for metrics, AI alerts, and user experience with other data for more informed custom support inquiries and enhanced operational performance reports.

Conviva MCP architecture: customer data sources and a customer agent connect through the MCP Server to Conviva analytics, sessions, and AI alerts

Customer agents reach Conviva analytics, sessions, and AI alerts through the Conviva MCP Server.

This guide provides examples using Claude Desktop, Cursor, and ChatGPT with an MCP Agent running locally within an organization. Conviva also supports using other Large Language Models (LLMs) and LLM clients that support the MCP Interface. Reach out to Conviva Support for details about LLM and LLM client support.
MCP FAQs

General Access

What data does Conviva MCP expose?

Conviva MCP exposes Video Streaming Insights (VSI) data, including:

  • Real-time metrics

  • Historical metrics

  • AI alerts (content and ad)

  • Viewer session data

  • KPI threshold definitions

  • Authorized c3 account access

All capabilities are available as MCP tools under the /mcp endpoint.

Authentication & Authorization

How do I authenticate to Conviva MCP?

Conviva MCP supports:

  • OAuth 2.0

  • API key authentication using HTTP Basic

How do autonomous agents authenticate?

Autonomous agents authenticate using a Conviva API key with HTTP Basic authentication.

The Authorization header must begin with:

Authorization: Basic <base64(client_id:client_secret)>

Conviva MCP systems can use Conviva's API keys via the Basic authentication scheme.

Where are Conviva API keys generated?

API keys are generated in the Pulse API Management page, where you obtain:

  • Client Id

  • Client Secret

Does Conviva MCP support OAuth?

Yes. Conviva MCP supports OAuth 2.0 authentication. Implementations should handle token refresh to prevent session expiration.

How do I rotate or revoke API keys or tokens?

General process:

  1. Generate a new API key or token.

  2. Update your MCP configuration.

  3. Verify the new connection.

  4. Revoke the old key or token.

For OAuth tokens, revocation is typically performed via the token revocation endpoint (POST /oauth/revoke). Credentials should be stored securely and rotated periodically.

Data & Tool Capabilities

Can MCP return real-time and historical metrics?

Yes. MCP tools support both real-time and historical VSI metrics.

Can MCP return per-session and per-viewer data?

Yes. MCP provides tools to:

  • List all sessions for a specific viewer within a time range

  • Retrieve aggregated quality metrics across sessions for a viewer

Can MCP retrieve AI alerts?

Yes. MCP tools allow you to:

  • List AI-detected quality alerts (content and ad)

  • Retrieve full alert details and severity timelines by ID

Can MCP retrieve KPI threshold definitions?

Yes. KPI thresholds for quality benchmarking can be retrieved using the appropriate MCP tool.

Can I list the c3 accounts my credentials have access to?

Yes. The get-authorized-c3-accounts tool lists all accessible c3 accounts.

This tool requires OAuth authentication.

Operational Considerations

What happens if rate limits are exceeded?

If requests are throttled, review the applicable API rate limits and implement request queuing with backoff.

Are there payload size limits?

Large tool inputs or outputs may be truncated or rejected. Ensure requests are appropriately scoped.

Validation & Schema Management

How do I validate MCP tool responses?

To validate responses:

  • Review published tool schemas.

  • Compare responses against expected input/output structures.

  • Verify the content array structure (each item contains a type field and associated data).

  • Ensure responses are valid JSON.

  • Properly handle error responses, which include isError: true and error details.

Testing tool calls with utilities such as curl or Postman before production integration is recommended.

Connect to Conviva MCP Server

The Conviva MCP server is available only as a remote HTTP service. No local installation is required or supported. Use the Conviva server URL and your credentials to connect directly from your MCP client.

Server URL: https://mcp.conviva.com/mcp

To obtain your API credentials, visit the API Management page in the Conviva Pulse dashboard.

For the latest release updates and changes, see What's New in Conviva MCP.

LLM Setups

MCP with Claude Desktop

For Claude Desktop access to Conviva MCP servers, add the Conviva MCP server settings to the claude_desktop_config.json file. Most users enable SSO access for authentication.

  1. Update the claude_desktop_config.json file in the \Claude folder and enable Okta SSO access. 

    {
      "mcpServers": {
        "conviva": {
          "command": "npx",
          "args": [
            "mcp-remote",
            "https://mcp.conviva.com/mcp"
          ]
         }
       }
    }

  2. In the Claude desktop, update the Settings > Connectors list to connect to the Conviva MCP servers.

    After server authentication prompts, the Claude chat window appears.

Node.js may be required to run the npm (Node Package Manager) npx command that configures the connectors with the Conviva MCP servers. See the MCP documentation.
Claude needs a complete restart to apply the configuration changes. To log out of an OAuth session if you are using the NodeJS mcp-remote package, in the command line clear the ~/.mcp-auth directory with rm -rf ~/.mcp-auth. In the Windows app, exit the application. Also, delete any files in the \Users\<user_name>\AppData\Local\npm-cache\_npx folder to refresh the authentication. Then, restart Claude.

Authentication

The Conviva MCP server uses OAuth to connect MCP clients according to the Authorization - Model Context Protocol. OAuth is more secure than using an API key because it allows more granular permissions and user-based authorization. When adding the Conviva MCP to a client, the MCP client opens an OAuth consent form which allows you to authorize the client to access your Conviva data.

If your MCP client doesn't support OAuth, pass your API credentials as a Basic Auth header. To generate your Base64-encoded credentials:

echo -n "your_client_id:your_client_secret" | base64

Then use the result as:

Authorization: Basic <YOUR_BASE64_CREDENTIALS>

To obtain your client_id and client_secret, visit the API Management page in the Conviva dashboard.

The get-authorized-c3-accounts tool is only available with OAuth authentication.

Building Autonomous Agents

To build agentic software that calls the Conviva MCP programmatically, pass your API credentials as a Basic Auth header directly in your HTTP requests:

curl https://mcp.conviva.com/mcp \
    -H "Content-Type: application/json" \
    -H "Authorization: Basic <YOUR_BASE64_CREDENTIALS>" \
    -d '{
        "jsonrpc": "2.0",
        "method": "tools/call",
        "params": {
        "name": "get-video-metrics-metadata",
        "arguments": {}
    },
    "id": 1
    }'