AI Pricing Hub

What ai-pricing-hub does

The ai-pricing-hub MCP server gives AI assistants structured pricing and comparison tools for language models and cloud compute. Its LLM data comes from the OptimToken catalogue, which tracks more than 250 models and is refreshed daily. Results can include price, Chatbot Arena ELO, efficiency, capabilities, self-hostability inferred from licensing, and whether the returned figures were verified.

The server is read-only and does not require credentials. It is intended for decisions such as choosing a model for a defined workload, comparing named models at a monthly volume, estimating request and monthly spend, or reviewing compute rates by provider and region.

How it works

The hosted MCP endpoint is available at https://optimtoken-mcp.optimnow.io/mcp. LLM tools first request data from the OptimToken API. If that source is unavailable, they may use OpenRouter data or an embedded snapshot. The compute comparison uses the pricing API first and an embedded snapshot as its fallback. Responses identify the source tier through a provenance object and expose whether prices were verified.

Cost calculations use the workload details supplied to the tool, including request volume, input and output token shape, cache hit rate, and batch eligibility. The catalogue also provides eight predefined workload profiles: Support Ticket, Knowledge Q&A, Meeting Summary, Marketing Content, Coding Task, Invoice Processing, Call Summary, and Agent Workflow.

Setup and configuration

For normal use, add the hosted endpoint as an HTTP MCP server. Claude Code can use:

claude mcp add --transport http optimtoken https://optimtoken-mcp.optimnow.io/mcp

Claude.ai or Claude Desktop can add the URL through Settings and custom connectors. Cursor, Windsurf, and VS Code can add an HTTP MCP server entry pointing to the same address. The URL must include /mcp and omit a trailing slash for the hosted widgets to render correctly.

Local development requires Node.js 24 or newer:

npm install
npm run dev
npm test
npm run build

The fallback catalogue can be regenerated with npm run refresh-fallback. No environment variables or API credentials are specified as required.

Tools and capabilities

• compare-llm-models filters the model catalogue by price, ELO, efficiency, capabilities, and self-hostability.
• recommend-llm-model ranks up to three models against budget, ELO, capability, and self-hosting constraints. If no model satisfies every constraint, it returns labelled near matches.
• compare-models-side-by-side compares two to four named models across eight workload profiles, including list and optimized cost at a selected monthly volume.
• estimate-llm-cost calculates per-request and monthly costs from custom workload assumptions.
• compare-compute-pricing compares instance rates from AWS, Azure, GCP, OCI, OVH, DigitalOcean, and Alibaba by region and category.

Limitations and notes

The ai-pricing-hub MCP server can fall back to uncorrected upstream or embedded prices. Lower-tier data may not include the catalogue's verified corrections, so check the returned provenance and dataAsOf value before relying on an embedded snapshot. The fallback data is refreshed manually and can become outdated.

Pricing comes from third-party sources and is provided without warranty. Verify figures against vendor pricing pages before committing spend. Nothing sent to the five tools is stored according to the provided documentation.