1
key for everything
Grab your API key
Use the same key from your InferenceSaver dashboard. No separate MCP credential to manage.
InferenceMCP
Give your AI client the whole model catalog.
One remote MCP server, on the same routing layer as our API. Point Claude, Cursor, or Windsurf at it and let the model pick the cheapest provider automatically.
What is MCP, really?
The Model Context Protocol is an open standard for connecting AI assistants to external tools and data. Instead of every app building its own integration with every service, an MCP server exposes a fixed set of tools once, and any MCP client — Claude Desktop, Claude Code, Cursor, Windsurf — can call them the same way.
Host
The AI app you're already using — Claude, Cursor, Windsurf. It runs the assistant and manages the connection.
Client
A piece inside the host that speaks MCP and keeps a session open with a specific server.
Server
InferenceSaver's MCP server. It exposes tools like list_models and chat_completion, backed by the same routing layer as our API.
What your assistant can do
Four tools, all gated behind your own API key. No anonymous access to models or spend.
list_modelsBrowse every model available on InferenceSaver, filtered by provider or capability, with live pricing.
chat_completionRun a chat completion through the routing layer. The cheapest available provider is picked automatically.
get_usagePull recent token usage for the authenticated account, straight from the source.
get_balanceCheck the current credit balance without leaving the assistant.
Connect in three steps
Same key, one URL, and you're routing through every model on InferenceSaver.
60s
typical setup time
Point your client at the server
Add https://mcp.inferencesaver.com as a remote MCP server in Claude, Cursor, Windsurf, or any MCP-compatible client.
0
extra credentials to manage
Authenticate with a Bearer token
Pass your API key as a Bearer token. Every tool call runs under your account, with your usage and your rates.
1
key for everything
Grab your API key
Use the same key from your InferenceSaver dashboard. No separate MCP credential to manage.
60s
typical setup time
Point your client at the server
Add https://mcp.inferencesaver.com as a remote MCP server in Claude, Cursor, Windsurf, or any MCP-compatible client.
0
extra credentials to manage
Authenticate with a Bearer token
Pass your API key as a Bearer token. Every tool call runs under your account, with your usage and your rates.
1
key for everything
Grab your API key
Use the same key from your InferenceSaver dashboard. No separate MCP credential to manage.
60s
typical setup time
Point your client at the server
Add https://mcp.inferencesaver.com as a remote MCP server in Claude, Cursor, Windsurf, or any MCP-compatible client.
0
extra credentials to manage
Authenticate with a Bearer token
Pass your API key as a Bearer token. Every tool call runs under your account, with your usage and your rates.
1
key for everything
Grab your API key
Use the same key from your InferenceSaver dashboard. No separate MCP credential to manage.
60s
typical setup time
Point your client at the server
Add https://mcp.inferencesaver.com as a remote MCP server in Claude, Cursor, Windsurf, or any MCP-compatible client.
0
extra credentials to manage
Authenticate with a Bearer token
Pass your API key as a Bearer token. Every tool call runs under your account, with your usage and your rates.
Questions about MCP
Do I need a separate key for MCP?+
No. Use the same API key from your InferenceSaver dashboard. Every tool call runs under your existing account, balance, and rate limits.
Which clients can connect?+
Any MCP-compatible client that supports remote Streamable HTTP servers: Claude Desktop, Claude Code, Cursor, and Windsurf all work out of the box.
Can it spend my credits without me knowing?+
Every tool call is gated behind your Bearer token and shows up in your usage logs immediately. There is no anonymous or shared access to models.
Is this the same routing layer as the API?+
Yes. mcp.inferencesaver.com sits on top of the exact same routing and pricing as api.inferencesaver.com. Same models, same rates, same account.
One key. Every model. Any MCP client.
Get an API key from the dashboard, drop it into your MCP config, and your assistant can route to the cheapest provider for any model — automatically.