mcpField
P
proxyllm-mcp

proxyllm-mcp

Local
proxyllm.aiAI & Agents0Updated 15w ago

About proxyllm-mcp

proxyllm-mcp is an MCP server for managing and querying the ProxyLLM OpenAI-compatible LLM gateway. It serves developers and agents with model data, key management, savings estimates, and an account signup flow.

Key features

  1. 01List gateway models
  2. 02Calculate subscription savings
  3. 03Inspect routing-key permissions
  4. 04Sign up and verify by email OTP
  5. 05Create keys and review usage

Use cases

  1. 01Compare gateway subscription costs with API spending
  2. 02Check which models and lanes a routing key permits
  3. 03Provision a routing key for provider connections
  4. 04Register an account through an agent

MCP support

It exposes ProxyLLM model, savings, key, account, and usage functions as tools for MCP clients.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"ai-proxyllm-proxyllm": {
"command": "npx",
"args": [
"-y",
"proxyllm-mcp"
],
"env": {
"PROXYLLM_ACCOUNT_TOKEN": "<PROXYLLM_ACCOUNT_TOKEN>",
"PROXYLLM_ROUTING_KEY": "<PROXYLLM_ROUTING_KEY>"
}
}
}
}
Run commandnpx -y proxyllm-mcp
Environment variables to set
PROXYLLM_ACCOUNT_TOKENSecretOptional sk_/rk_ account token for management tools. Create one autonomously via the proxyllm_signup tool or https://proxyllm.ai/auth.md
PROXYLLM_ROUTING_KEYSecretOptional pllm_ routing key for key introspection and key-scoped model lists.

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmproxyllm-mcp1.0.2

Tools

9

Saved tool list · updated 3w ago

ParameterTypeRequiredDescription
monthly_openai_bill_usdnumberYes

Details

From the project README

proxyllm-mcp

MCP server for ProxyLLM, the OpenAI-compatible LLM gateway. It gives any MCP client (Claude Code, Claude Desktop, Cursor, Cline, Windsurf, OpenClaw) live model catalogs, plan-savings math, routing-key introspection and provisioning, and an account signup flow an agent can complete on its own: email OTP in, sk_ account token out, HTTP 402 with a checkout link until a human activates the membership.

Install

Claude Code:

claude mcp add proxyllm -- npx -y proxyllm-mcp

Cursor, Cline, Claude Desktop, or any client that takes a JSON server entry:

{
  "mcpServers": {
    "proxyllm": {
      "command": "npx",
      "args": ["-y", "proxyllm-mcp"]
    }
  }
}

No configuration is required to start. Two optional environment variables save passing tokens per call:

  • PROXYLLM_ACCOUNT_TOKEN: an sk_ (read+write) or rk_ (read-only) account token for the management tools.
  • PROXYLLM_ROUTING_KEY: a pllm_ routing key for key introspection and key-scoped model lists.

Tools

Free, no account needed:

  • proxyllm_savings_calculator: monthly OpenAI API bill in, subscription tier + flat total + monthly savings out.
  • proxyllm_list_models: OpenAI-shaped model list from the gateway.
  • proxyllm_introspect_key: what a routing key can do (lanes, models, budget, whether a Codex subscription backs it).

Account flow (an agent can run this end to end):

  • proxyllm_signup: emails a 6-digit code to the operator's inbox. No captcha.
  • proxyllm_verify_signup: exchanges the code for the sk_ account token (returned once).
  • proxyllm_account: account state; while unpaid it carries the checkout link and a message written to be relayed to the operator. Poll until plan is "pro".

Provisioning (paid accounts):

  • proxyllm_create_routing_key: mints a pllm_ key, optionally wiring its provider chain (Codex subscription seats, self-hosted Claude Code bridges, metered API keys) in the same call.
  • proxyllm_list_routing_keys: keys, plus the wireable lane ids with include_lanes.
  • proxyllm_usage: 30-day usage summary.

Unpaid accounts get 402 payment_required on management tools by design; the body names the price ($129/mo flat), the checkout URL, and the 48-hour removal deadline for never-activated accounts. Error bodies are returned verbatim because they contain the next step.

Links

MIT license.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.