Skip to main content
Glama

Etuki FinOps MCP Server (etuki-finops-mcp)

MCP Specification Glama MCP Server License: MIT

etuki-finops-mcp is an official Model Context Protocol (MCP) Server built on the @modelcontextprotocol/sdk standard (supporting both STDIO and Streamable HTTP / SSE transports at https://www.etuki.com/api/v1/mcp).

It equips MCP clients (Claude Desktop, Claude Code, Cursor, Windsurf, Cline, ZCode, and Gemini CLI) with 9 deterministic MCP Tools, 2 MCP Resources, and 2 MCP Prompts for real-time LLM API pricing lookup across 563+ models, 0.1x/0.025x prompt-cache breakpoint locking, relay downgrade detection, and SHA-256 billing receipt reconciliation.


πŸ—οΈ MCP Server Implementation in This Repository


Related MCP server: ai-cost-calculator-api

πŸ› οΈ Exposed MCP Tools (tools/list β€” 9 Read-Only Deterministic Tools)

All 9 MCP tools implement strict JSON Schema inputSchema, outputSchema, and MCP security annotations (readOnlyHint: true, destructiveHint: false, idempotentHint: true, openWorldHint: false):

#

MCP Tool Name

Description

Required Parameters

1

discover_and_probe_account_models

Probes available & free-tier models in the user's connected API account to offload non-architectural file scans.

primary_model (string)

2

auto_route_task_by_quality_tier

Computes a 6-turn Pareto routing plan (4 scan turns + 2 flagship synthesis turns) with dual cash vs full-economic accounting.

task_instruction (string), context_tokens (number)

3

pin_single_model_prefix_cache

Calculates 10-turn prompt cache savings and places 4 cache_control: ephemeral breakpoints without changing the user's active model.

model (string), prompt_tokens (number)

4

compress_and_fold_context

Folds repetitive build/terminal logs by up to 90% while preserving 100% of AST/compiler error lines.

raw_log_text (string)

5

detect_model_downgrade_and_substitution

Evaluates channel risk for silent model downgrade or quantization on third-party API relays.

model (string), burned_usd (number), remaining_usd (number)

6

reconcile_usage_receipt

Audits an LLM API usage receipt against official 1x vendor rate cards to detect relay markups and unpinned cache waste.

model (string), prompt_tokens (number), actual_billed_usd (number)

7

check_context_cliff

Detects >128K / >200K context window surcharge cliffs and recommends optimal sliding-window truncation.

model (string), prompt_tokens (number)

8

list_supported_models (etuki_search_models)

Queries the 563-entity Merkle-verified LLM & cloud rate card catalog (input_usd_1m, output_usd_1m, cache_read_usd_1m, context_window).

query (string)

9

etuki_get_audit_evidence

Retrieves the RS256 cryptographic signature envelope, 12-field Merkle inclusion proofs, and double-entry ledger state.

verify_chain (boolean)


πŸ“š Exposed MCP Resources (resources/list)

  1. etuki://catalog/leaves-563 (application/json): 563-Entity 12-Field Merkle Rate Card & Model Compatibility Registry.

  2. etuki://ledger/zcode-152-audit (application/json): SHA-256 Reconciled Production Billing Audit Ledger.


πŸ’¬ Exposed MCP Prompts (prompts/list)

  1. pre_flight_cost_saver_workflow: Pre-flight cost optimization workflow combining model discovery, 6-turn routing, and prefix cache pinning.

  2. sha256_bill_forensic_audit: Forensic audit prompt for verifying API usage receipts against official 1x vendor pricing.


⚑ Quick Start & MCP Client Configuration

Option A: Claude Desktop (claude_desktop_config.json) or Cursor (.cursor/mcp.json) β€” Remote Streamable HTTP

{
  "mcpServers": {
    "etuki-finops-mcp": {
      "url": "https://www.etuki.com/api/v1/mcp"
    }
  }
}

Option B: Local STDIO Execution via Node.js (index.mjs)

{
  "mcpServers": {
    "etuki-finops-mcp": {
      "command": "node",
      "args": ["/path/to/etuki-gate/index.mjs"]
    }
  }
}
# Run STDIO MCP Server locally:
npm install
node index.mjs

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI cost calculation, comparison, and optimization across major providers like Anthropic, OpenAI, Google, Meta, and Mistral. Supports cost estimation, budget-aware model finding, and token estimation through a simple API and MCP integration.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Live LLM API pricing: current token prices, model comparisons, cheapest-model lookups, and The LLM Price Index for 150+ models across 20+ providers, re-verified daily. No API key required.
      4
      1
      MIT