#cost-optimization

Multi-Agent Code Review Cost Benchmark 2026: Librarian vs Prompt Cache

Get notifications about leaks that might be bleeding your tokens. Based on real-time analysis on your coding agent sessions.

See where LLM money goes, why, and how to cut waste. Local-first FinOps + spend intelligence for Claude Code & the Anthropic API. Pure stdlib, zero deps.

A Claude Code plugin that shows exactly where your AI coding session wasted tokens — and how to fix it. Flags waste, never quality. Runs locally.

Precision for your AI spend — see exactly where your Claude Code money goes. Local dashboard + optimization skill (caliper.run)

Auto-route each Claude Code session to Haiku, Sonnet, Opus or Fable from its first prompt. Classified locally, zero tokens.

Reliability layer for AI coding agents: shared memory that survives sessions, blast-radius checks before edits, verification gates after — one config for 10 tools, zero runtime deps.

Token usage tracker for multi-agent AI sessions

When Jev meets LLM — Transparent proxy that makes AI coding agents 3× faster and 80% cheaper. Cut response times from 2.4s to 890ms with Line J architecture.

Snowflake Cortex demo: 11-lever cost+quality optimization framework for legal-PDF document AI workloads

Flagship free FastAPI starter for LLM routing/cost visibility; $29 Gumroad kit adds deployment, hardening, routing, and verification guides.

PECS: 基于 LangGraph 的四角色多智能体任务求解框架（Planner/Executor/Critic/Synthesizer）。WebShop 真实环境 +25pp (25% vs 0%)；GAIA 官方 53 题 26.4% vs ReAct 24.5%（McNemar 不显著）。含 AST 沙箱、50000 token 硬预算、FastAPI 限流/混沌/CI/Prometheus。

Picks the model tier for a delegated Claude Code task before it spawns. Opt-in per repo, reversible in one command, fails safe. v0.1 is a lookup table on purpose.

Automatic model routing + live agent dashboard for Claude Code. Haiku for lookups, Sonnet for daily work, Opus to orchestrate - watch every agent, model and dollar in real time.

Semantic caching layer for LLM & agent tool calls — matches near-duplicate prompts by embedding similarity to cut redundant API spend and latency.

Cost-aware AI agent orchestration for coding projects: strong models decide, cheaper models execute bounded work, and one lead owns final acceptance.

Zero-Overhead Edge Circuit Breaker & Token Runaway Protection for LLM APIs (GPT-4o, Claude 3.5, Gemini).

MCP server for the GONKA network: run cheap LLM inference through the server (free trial or your own key), get multi-model second opinions, and compare live prices — for any AI agent.

A controller that learns how a tool-using agent should spend compute at each step (WIDER, DEEPER, DECOMPOSE, STOP, or ESCALATE to a stronger model) and improves its own policy from logged traces. Judged on cost per solved task with paired statistics; ships trained policies and an offline embed API (pip install, no key needed).

Audits real Claude Code usage for effort/model waste — reuses ccusage for raw data, adds the judgment layer it can't provide.

Meerada — the live AI model exchange (every model measured hourly, priced by cost per verified task) + LLManager: import your Claude/ChatGPT history and switch models mid-conversation. Free, open beta.

Claude Code hooks that measure live context, warn as it grows, and force a detailed handoff note before a long chat gets expensive.

Where an agent loop's wall-clock actually goes: measured TPS/transcript-tax benchmark, with retractions. Data + analysis scripts + figures.

BYOK OpenAI-compatible LLM router with difficulty classification, prompt caching, and quality-aware cost optimization.

Confidence-gated model routing for coding agents — run work at the cheapest model that can pass verification. SKILL.md, MIT.