MCP

Omniconvert MCP Server exposes experiments and analytics data through the Model Context Protocol (MCP).

Compatibility:

• MCP spec: 2025-06-18

• JSON-RPC: 2.0

• Supported transport: Streamable HTTP

MCP entry point: https://web.omniconvert.com/mcp

Custom connectors for MCP clients

• Claude AI connector

• Chat GPT connector

• API Key clients (Cursor, Claude Code, Claude Desktop)

Authentication

The server accepts two auth schemes on the standard Authorization header:

• ApiKey — Authorization: ApiKey <token>. Recommended for programmatic clients (Cursor, Claude Code CLI, Claude Desktop via bridge, custom integrations). See the API Key clients page for how to generate a key and wire it into a config file.

• OAuth 2.1 Bearer — Authorization: Bearer <token>, with dynamic client registration. Used automatically by the Claude AI and ChatGPT custom connectors.

Tokens are scoped to the Omniconvert account that owns them. A token can read experiments, variations, goals, segments and reports for any website attached to that account; it cannot reach resources owned by a different account.

Tools and Resources

Available tools are:

• init – retrieve initial context for experiments analysis

• find_website – find a website by its ID or URL

• list_websites – list all available websites

• list_experiments – get the list of available experiments

• read_experiment – get experiment data by its URI

• read_variation – get variation data by its URI

• read_goal – get goal data by its URI

• read_segment– get segment data by its URI

• list_experiments_performance_overview – bulk performance overview for up to 10 experiments (users, conversion rate, revenue, leads, survey respondents) over a date range

• get_experiment_performance_report – get performance report for an experiment

• get_experiment_performance_chart_series – get performance chart series for an experiment

Available resources:

• Experiment

• Variation

• Goal

• Segment