PlugWiki

OpenRouter MCP: look up models, prices and credits from your coding assistant

OfficialRemote
Gethttps://mcp.openrouter.ai/mcpDocs

Lets your coding assistant look up OpenRouter models, prices, credits and rankings, and send test messages while you build.

This is OpenRouter's own server, and it runs on OpenRouter's side, so there is nothing to install. It helps Claude Code, Codex, Cursor and other assistants answer questions like which model is best for a job, what it costs and how much credit you have left. You add one address and approve a sign-in in your browser. It is a helper for building, not a way to run models from your app, which should keep calling the OpenRouter API.

Good for

  • Developers choosing which model to use, based on live prices and benchmarks instead of the assistant's memory
  • Builders checking their OpenRouter credit and what a test call cost without opening the dashboard
  • Engineers comparing answers from several models on the same prompt
  • Anyone asking how to do something in OpenRouter's docs from inside their editor

Works in

5 supported · 9 should work

Tap an app to see how to set it up and any known problems.

Install OpenRouter MCP

Claude Code

  1. 1

    Copy this prompt.

    Add the OpenRouter MCP server to Claude Code. Run this command:
    
    claude mcp add --transport http openrouter https://mcp.openrouter.ai/mcp
    claude mcp login openrouter
    
    Then check that it loaded and tell me if I need to sign in or restart.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires OpenRouter account.

When you sign in, OpenRouter asks you to approve a separate key just for this connection. It starts with a $10 spending cap, which you can change on that screen, and it expires within 7 days; sign in again when tools start failing with an auth error. Only test messages and image generation spend credits. You can also sign in from inside a session: run /mcp, pick openrouter and choose Authenticate.

From the vendor's docs. Source.

Tools

  • Searches OpenRouter's live model list by price, context, benchmarks and more

    Searches the live catalog of models on OpenRouter, with sorting and filters for price, context length, benchmark scores, tool-calling success, provider, input and output types, and data policy.

    Try: “Find models under $1 per million tokens with at least 200K context”

    Network · list-models

  • Shows full details for one model

    Returns everything OpenRouter knows about one model, such as its pricing, context size and supported features.

    Try: “Tell me everything about anthropic/claude-opus”

    Network · get-model

  • Shows which providers serve a model and their price and speed

    Lists the providers that serve a model, with each one's price, latency, throughput and data policy.

    Try: “Which providers serve DeepSeek v4 Pro, and which is fastest?”

    Network · list-model-endpoints

  • Shows third-party quality scores for models

    Returns quality scores from Artificial Analysis (intelligence, coding and agentic indexes) and Design Arena (head-to-head ratings and win rates).

    Try: “Which model tops the Artificial Analysis intelligence index?”

    Network · list-benchmarks

  • Shows which models are most used and trending

    Lists the models that are most used and rising on OpenRouter, by token volume.

    Try: “Which models are trending on OpenRouter this week?”

    Network · list-daily-model-rankings

  • Shows what people use OpenRouter for and top models per task

    Breaks down OpenRouter traffic by task type, such as code generation, web search or summarization, with the top models for each.

    Try: “Which models are used most for code generation?”

    Network · list-task-classifications

  • Sends a test message to any model and returns its reply (billed)

    Sends a prompt to any OpenRouter model and returns the answer, to test a prompt or compare models. This is a paid call charged to your credits.

    Try: “Send a test message to GPT-5.5 and tell me what it cost”

    Write · send-message

  • Creates an image from a text prompt (billed)

    Generates an image from your text and shows it in the chat. This is a paid call charged to your credits.

    Try: “Generate a simple logo idea for a coffee shop called Bean There”

    Write · generate-image

  • Shows how much OpenRouter credit you have left

    Returns the credit remaining on your OpenRouter account.

    Try: “How much credit do I have left on OpenRouter?”

    Network · get-credits

  • Shows the cost, tokens and provider for one past call

    Looks up one generation by its ID and returns what it cost, how many tokens it used and which provider served it.

    Try: “What did that last test message cost?”

    Network · get-generation

  • Searches OpenRouter's documentation

    Searches the full OpenRouter docs to answer how-to questions.

    Try: “How do I pin a model to a specific provider, e.g. Bedrock?”

    Network · search-docs

  • Lists the providers you can route requests to

    Lists the model providers available on OpenRouter, for setting routing preferences.

    Try: “Which providers can I route requests to?”

    Network · list-providers

From the official docs

Pros and cons

Good

  • Made and hosted by OpenRouter, nothing to install.
  • Sign in through your browser; no key to copy or paste.
  • Each connection gets its own key with a 7-day life and a $10 starting cap.
  • Model advice comes from live prices, rankings and benchmarks, not the assistant's memory.
  • Most tools are free lookups.
  • Shows what each test call cost and which provider served it.

Watch out

  • Not meant for running models from your app.
  • The key expires within 7 days, so you sign in again each week.
  • Test messages and images are billed to your credits.
  • Not in Claude's connector directory, so you add it by hand.
  • Some organizations block custom connectors in Claude.

Permissions

low risk · 3 things to watch

Low risk

The main risk is cost, because test messages and image generation are billed to your OpenRouter credits.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • Test messages and images spend your credits

    Sending a test message or generating an image is billed to your OpenRouter account. Keep the $10 cap, or set a lower one on the approval screen. openrouter.ai

  • Each connection gets its own short-lived key

    The key expires within 7 days and you can revoke it in the keys dashboard, so a leaked key stops working on its own. openrouter.ai

  • Your text leaves only when you send it

    Code or prompts go to OpenRouter only through test messages, image prompts or feedback. openrouter.ai

Security advisories

No published security advisories as of Nov 1, 2026.

Technical details
Turning on read-onlyNo read-only switch. Two tools make billable model calls and one sends feedback; the spend cap on the connection's key limits cost.
Talks tomcp.openrouter.ai
Sign-in and keysA dedicated OpenRouter API key issued at sign-in, held by your AI app, expiring within 7 days.

Alternatives

Use OpenRouter's own server to research models, prices and credits while you build. To have your assistant actually chat with OpenRouter models or handle images, pick a community server such as openrouter-mcp-multimodal.

ServerOfficialRunsSign-inToolsGitHub stars
For OpenRouter
OpenRouterThis pageOfficialHostedAccount17–
openrouter-deep-research-mcpCommunity–––55
Other tools for the same job
openrouter-mcp-multimodalCommunity–––102

Questions

Is the OpenRouter MCP server free?

Mostly. Connecting and the lookup tools cost nothing, but test messages and image generation are billed to your OpenRouter credits.

What is the OpenRouter MCP server URL?

https://mcp.openrouter.ai/mcp. You add it to your AI app and sign in through your browser.

Why does it stop working after a few days?

The key it creates expires within 7 days, and your account's key policy can make that shorter. Run your app's sign-in step again to get a new one.

Can I use it to run models in my app?

No. It is a helper for while you build. Your app should keep calling the OpenRouter API directly.

How much can it spend?

Each connection's key starts with a $10 spending cap, which you can change on the approval screen. Only test messages and image generation spend credits.

How do I disconnect it?

Revoke the key named OpenRouter MCP followed by your app's name in the keys page of your OpenRouter dashboard. It also stops on its own when it expires.

How do I add it to Claude Desktop?

It is not in Claude's connector directory, so add it as a custom connector with the address above and leave the sign-in fields blank. The install box above has the steps.

Does it see my code?

Only what you send. Text goes to OpenRouter when the assistant sends a test message, an image prompt or feedback.

Examples

  • Pick a model for a task from live benchmarks and rankings

    What's the best model for coding right now?

    Source
  • Check how much credit is left

    How much credit do I have left on OpenRouter?

    Source
  • Test a model and see what the call cost

    Send a test message to GPT-5.5 and tell me what it cost.

    Source
  • Compare several models on one prompt

    Compare the answers for the same prompt across Opus 4.8, DeepSeek v4 Pro, Gemini 3.5 Flash, and GLM-5.2.

    Source
  • Find out how to route to one provider

    How do I pin a model to a specific provider, e.g. Bedrock?

    Source

Information

updated Nov 1, 2026

Publisher
Official
Endpoint
https://mcp.openrouter.ai/mcp
Transport
Streamable HTTP
Auth
OAuth

Docs

Other OpenRouter MCP servers

Sources

Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.

Show all 2
  1. MCP · openrouter.ai
  2. GitHub · github.com

Facts from the official MCP registry, GitHub, npm, PyPI and the Docker catalog. Updated Nov 1, 2026.