MCP Server

Route smarter.
Spend less.

WhichModel is an MCP server your agents call to pick the right AI model for any task. Stop overpaying for GPT-4 on jobs that Haiku handles perfectly.

// Add to your agent config
"mcpServers": { "whichmodel": { "url": "https://mcp.whichmodel.dev/mcp" } }

Tools

Four tools. Zero guesswork.

Add WhichModel to your agent once — it calls the right tool automatically.

recommend_model

Get a cost-optimised model recommendation for a task type, complexity level, and budget ceiling.

compare_models

Head-to-head comparison of 2–5 models with optional volume cost projections.

get_pricing

Raw pricing data lookup with filters by model, provider, price, and capabilities.

check_price_changes

See what model pricing has changed since a given date. Keep your cost estimates fresh.

Pricing

One saved routing decision pays for weeks of the Developer plan.

Free

^$0

1,000 requests / month

What counts as a request?

Each MCP tool call counts as one request — recommend_model, compare_models, get_pricing, or check_price_changes.
Do I need an API key to start?

No. The free tier works immediately via IP-based rate limiting — just point your agent at https://mcp.whichmodel.dev/mcp. An API key removes IP restrictions and unlocks the usage dashboard.
What happens when I hit my monthly limit?

You'll receive a 429 response with a Retry-After header. Upgrade anytime to continue without interruption.
Can my agent buy a plan autonomously?

Not yet — plans are managed by the developer or operator. Credits-based billing for autonomous agents is on the roadmap.
Which AI providers are covered?

Anthropic, OpenAI, Google, Meta, Mistral, Cohere, and more. The model database is updated whenever major providers publish pricing changes.