WhichModel

whichmodel/whichmodel-mcp

This server connects Claude to WhichModel's model recommendation API, which helps autonomous agents pick the right LLM for each task based on live pricing data. You ask it a question about what model to use, and it returns a single confident recommendation optimized for cost. Useful when you're building agents that need to route requests across multiple models and want to minimize API costs without manually tracking pricing changes. The service limits free queries, so you'll likely need an API key for production use. Exposes streamable HTTP transport, meaning it works with Claude Desktop and other MCP clients that support remote connections.

tools: 6
transport: HTTP
last updated: Apr 7, 2026

View on GitHub

Tools

6 tools

Public tool metadata for what this MCP can expose to an agent.

recommend_modelGet a cost-optimised model recommendation for a specific task. Describe what you need to do and get back the best model with cost estimate, reasoning, and alternatives.6 params

Get a cost-optimised model recommendation for a specific task. Describe what you need to do and get back the best model with cost estimate, reasoning, and alternatives.

Parameters* required

budget_per_callnumber

Maximum spend in USD for this single call

complexitystring

Task complexity: low, medium, or highone of low · medium · highdefault: medium

estimated_input_tokensinteger

Estimated input size in tokens

estimated_output_tokensinteger

Estimated output size in tokens

requirementsobject

Additional requirements for the model

task_typestring

The type of task you need a model forone of chat · code_generation · code_review · summarisation · translation · data_extraction

compare_modelsHead-to-head comparison of 2-5 specific models. Compare pricing, capabilities, quality tiers, and optionally project costs based on expected usage volume.3 params

Head-to-head comparison of 2-5 specific models. Compare pricing, capabilities, quality tiers, and optionally project costs based on expected usage volume.

Parameters* required

modelsarray

Model IDs to compare, e.g. ["anthropic/claude-sonnet-4", "openai/gpt-4.1"]

task_typestring

Task type for context-aware comparisonone of chat · code_generation · code_review · summarisation · translation · data_extraction

volumeobject

Expected usage volume for cost projections

get_pricingRaw pricing data lookup for LLM models. Filter by model, provider, price ceiling, and capabilities. Returns structured pricing data for programmatic consumption.7 params

Raw pricing data lookup for LLM models. Filter by model, provider, price ceiling, and capabilities. Returns structured pricing data for programmatic consumption.

Parameters* required

capabilitiesarray

Required capabilities to filter by

include_deprecatedboolean

Include deprecated and sunset models in results (default false)default: false

limitinteger

Maximum number of results to return (1-100, default 20)default: 20

max_input_pricenumber

Maximum input price per million tokens in USD

min_context_windowinteger

Minimum context window size in tokens

model_idstring

Specific model ID, e.g. "anthropic/claude-sonnet-4"

providerstring

Filter to models from this provider, e.g. "anthropic"

check_price_changesCheck what model pricing has changed since a given date. Useful for monitoring cost changes and spotting new models or deprecations.3 params

Check what model pricing has changed since a given date. Useful for monitoring cost changes and spotting new models or deprecations.

Parameters* required

model_idstring

Filter to a specific model

providerstring

Filter to a specific provider

sincestring

ISO date to check changes from, e.g. "2026-04-01"

estimate_costEstimate the cost of a specific workload for a given model. Returns cost per call, daily and monthly projections, and a comparison to the cheapest alternative with equivalent capabilities.4 params

Estimate the cost of a specific workload for a given model. Returns cost per call, daily and monthly projections, and a comparison to the cheapest alternative with equivalent capabilities.

Parameters* required

calls_per_dayinteger

Expected number of calls per day (for daily/monthly projections)

input_tokensinteger

Number of input tokens per call

model_idstring

Model ID to estimate cost for, e.g. "anthropic/claude-sonnet-4"

output_tokensinteger

Number of output tokens per call

find_cheapest_capableFind the cheapest models that meet specific capability requirements. Useful when you have hard constraints (e.g. must support tool_calling + vision) and want the most cost-effective option.3 params

Find the cheapest models that meet specific capability requirements. Useful when you have hard constraints (e.g. must support tool_calling + vision) and want the most cost-effective option.

Parameters* required

min_context_windowinteger

Minimum context window size in tokens, e.g. 128000

quality_floorstring

Minimum quality tier: "low" (budget+), "medium" (standard+), "high" (premium+), "frontier" (frontier only)one of low · medium · high · frontier

required_capabilitiesarray

Capabilities the model must support, e.g. ["tool_calling", "json_output", "vision"]

Readme

Details

Repository: github.com/whichmodel/whichmodel-mcp
Website: whichmodel.dev
Category: AI & LLM Tools
Transport: HTTP
Registry: MCP registry, active

Related AI & LLM Tools MCP servers

View all

Postboltwhitemarmot/postbolt-mcpPublish to 13+ social media platforms from any AI agent via MCP.0stars

Xaip Mcp Serverxkumakichi/xaip-protocolAI agents get on-chain identity, credentials, reputation, escrow, and persistent memory on XRPL.0stars

mindmap.iomohganji/mindmapio-mcpCreate mindmaps of any topic from Claude, Cursor or any AI agent. Every node can be an AI chat.0stars

Opseraopseraengineering/opsera-aiOpsera AI Agent MCP server providing authenticated DevOps automation over SSE0stars

Mcpscispot-repo/scispot-mcp-serverTurn any LLM into your lab assistant: search samples, track experiments, analyze data with AI.0stars

Mcpsiteglassio/siteglass-mcpArchive any URL to a shareable rrweb/pdf/png snapshot + agent-driven web QA (scan, run E2E flows).0stars