Home
MCP server

Use AgentKernel in Claude, ChatGPT and Cursor

With the AgentKernel MCP server your AI assistant queries the live data directly, for example: “Which open model is strongest for coding?” or “What does Claude Opus 5.5 cost at 5M tokens a month?”. Free, no account, read-only.

Server URL (Streamable HTTP)

https://www.agentkernel.si/mcp

Setup

Claude (web and desktop)

Settings → Connectors → add a custom connector, name “AgentKernel”, URL as above.

Claude Code

In the terminal:

claude mcp add --transport http agentkernel https://www.agentkernel.si/mcp

Cursor

Add to ~/.cursor/mcp.json:

{ "mcpServers": { "agentkernel": { "url": "https://www.agentkernel.si/mcp" } } }

VS Code

Add to .vscode/mcp.json:

{ "servers": { "agentkernel": { "type": "http", "url": "https://www.agentkernel.si/mcp" } } }

ChatGPT

In developer mode, add a custom MCP server with the URL above under Settings → Connectors (availability depends on your plan).

Tools

best_for_task: Best models per task, ranked by performance only: text, coding, agents, websites, images, video, speech and more.
recommend_models: Recommendations by use case and monthly budget using the Fit Score, optionally with privacy and licence requirements.
get_model: One model’s facts: price, context, ranks, licence, privacy, cost and alternatives.
find_models: Search and filter the catalog by name, provider, licence and privacy.
estimate_cost: Monthly cost for up to five models at your token volume.
cheaper_alternative: A cheaper alternative at most five index points below your current model.

Example questions

Which model is best for text-to-video right now, and which of them can I book via API?
Recommend a coding model for $25 a month that runs with zero data retention in the EU.
Is there a cheaper alternative to GPT-6.1 Sol at 10M input tokens a month?
Compare the cost of Claude Opus 5.5, Gemini 3.8 Flash and GLM 5.3 at 2M tokens.

Read-only, no sign-in. AgentKernel does not store your requests; the host keeps standard access logs. Data is synced hourly. Please credit AgentKernel when you pass the figures on.