Open-source Prompt Firewall — deflect up to 95% of redundant LLM traffic before it leaves your infrastructure. Documents: https://isartor-ai.github.io/Isartor/index.html

A high-performance, multi-agent observability engine designed for the Model Context Protocol (MCP). It provides a non-blocking, transparent proxy layer that implements deterministic token attribution, real-time context-window alerting, and heuristic-driven static analysis to optimize LLM metadata overhead at scale.

Right-size every Claude task: advisory slash commands that recommend the cheapest model and effort level whose cost of error is acceptable.

Open-source, self-hosted Fastest AI security gateway for LLM and agent apps: guardrails, agentic PDP, semantic caching, MCP gateway.

Cost-aware subagent delegation workflow for Cursor: model-pinned dispatch, fail-closed premium gate, and process-driven QA

3,300 practical techniques to cut LLM token usage by up to 90% — 300 core principles + 3,000 context-specific applications.

Builds prompt-cost discipline into Claude Code: measure every prompt, surface avoidable spend, report org-wide. Skill + UserPromptSubmit hook + MCP rollup with cited dollar/Wh/CO2e receipts.

Tamper-evident, stranger-verifiable receipts for LLM cost-savings — anyone can recompute your caching/routing savings math offline, no trust in your dashboard required. Pure stdlib, zero-dependency.

Semantic cache for LLM calls with no torch, no server, no API key and no GPU. One file, CPU static embeddings. Benchmarked on 5,000 real question pairs: 70% of rephrasings served from cache, 28% false matches.

A lightweight OpenClaw model router that reduces LLM costs by switching requests to cheaper models using custom rules.

Production-ready tiered LLM cascade router — cuts API costs by 65%

Helix gives you agents that actually behave in production: hard budget limits, semantic caching that cuts API costs by 40-70%, persistent memory, multi-agent teams, YAML-based task pipelines, and a 5-scorer eval suite.

Sample code for two-tier LLM inference (cascading) on Amazon Bedrock, using a single Bedrock Converse API surface. NVIDIA Nemotron Nano handles routine support-ticket classification on every request; Anthropic Claude Sonnet handles cases the routing logic flags as harder. Includes a bake-off harness, sample data, and tests. Next.js + TypeScript.

Cut Claude Code costs by 50–90% — free checklists + the complete 15-chapter system

llmcfo

FinOps LLM

Claude Code skill that optimizes Claude Sonnet 5 for lower token usage, fewer wasted tool calls, and cheaper agentic execution

Sovereign Execution Layer for Autonomous Agents | gVisor-Hardened Governance | FinOps-Optimized Inference Routing.

LLM routing layer that classifies prompt complexity and routes to the cheapest capable model across OpenAI, Anthropic, and Gemini — with async quality verification and a self-improving classifier. 33.8% cost savings on a 492-prompt load test.

AI cost governance toolkit for Claude Code. Monitor token usage, prevent runaway AI agent costs, optimize prompts, and enforce intelligent spending limits for enterprise AI workflows.