Cost Optimization Python Packages

skypilot

The AI Compute Platform for frontier teams. SkyPilot turns fragmented AI compute into one AI supercomputer, so frontier AI teams build custom intelligence faster.

metaflow

Build, Manage and Deploy AI/ML Systems

skypilot-nightly

The AI Compute Platform for frontier teams. SkyPilot turns fragmented AI compute into one AI supercomputer, so frontier AI teams build custom intelligence faster.

metaflow-stubs

Build, Manage and Deploy AI/ML Systems

trainy-skypilot-nightly

The AI Compute Platform for frontier teams. SkyPilot turns fragmented AI compute into one AI supercomputer, so frontier AI teams build custom intelligence faster.

xrouter-llm

A prompt-aware LLM router that predicts which models can complete each request, then selects the cheapest capable one: 53.2% lower cost and +1.9 pts completion on our tested dataset.

pyoptex

Flexible and accessible design of experiments in Python. Provides industry with an easy package to create designs based with limited expert knowledge. Provides researchers with the ability to easily create new criteria and design structures.

distil-llm

Compression with a quality contract — cache-aware, causally-pruned LLM context compression for agentic runtimes, certified non-inferior across 7 domains. Works with any SDK.

llm-routing

Universal LLM router for AI coding tools. Works with Claude Code, Cursor, Codex, Gemini CLI, Copilot and more. Free-first fallback chain cuts costs 35-80%.

ondine

Batch-process tabular data with any LLM. pandas & Polars in, structured columns out. Checkpointing, budget limits, and 100+ providers built in.

localharness

An open-source, coding agent harness designed specifically for local LLMs. Designed as everything a user needs to jumpstart their own harness (detailed event steams, logs, and performance data) or quickly replace closed source one (featuring out of the box plugins for web search, image gen, memory, auto research and more in dev).

attune-ai

Persistent memory and receipt-verified workflows for Claude Code — plugin, MCP server, and spec-driven dev framework in one pip install.

finops-mcp

The headless FinOps intelligence layer: API-first, agent-first cost control for cloud + AI. Connect your cloud stack in Claude, Cursor, or the terminal, and gate what agents do to your infra before they act. Propose-only. Open source, local-first.

spendguard-mcp

Spend governor for AI agents on data warehouses — pre-flight dollar-cost estimates, hard budgets, and a self-calibrating spend ledger. BigQuery, Snowflake, Databricks.

omnicache-proxy

⚡ Zero-latency semantic caching, agent loop accelerator & cost optimization gateway for LLMs. Slash OpenAI & Anthropic bills by 40%-75%.

aws-analyzer

Readable reports on your AWS data from a SageMaker or Jupyter notebook: S3 buckets, DynamoDB tables, Bedrock knowledge bases and the SageMaker notebook itself. What's there, what it costs, and what to do next. Read-only, boto3 only; one file per service, or pip install aws-analyzer.

cloudcosttree

Estimate AWS infrastructure costs in a hierarchical tree — before you apply.

modelhop

Intelligent LLM router that saves 60-90% on API costs by hopping to the cheapest capable model

piqc

Free, read-only GPU waste scanner for Kubernetes inference clusters, by Paralleliq. Finds idle GPUs, oversized GPU tiers and wasted capacity in vLLM deployments.

claude-code-llm-router

Universal LLM router for AI coding tools. Works with Claude Code, Cursor, Codex, Gemini CLI, Copilot and more. Free-first fallback chain cuts costs 35-80%.

cost-guard-mcp

Pre-flight query cost and result-size guardrails for AI agents across BigQuery, Snowflake, and Databricks

cafitac-hermit-agent

Quiet MCP executor for Claude Code and Codex. Offload edits, tests, refactors, and commits to cheaper local or flat-rate models.

cloudcostwise

Find AWS waste from your terminal or your AI assistant (CLI + MCP server). Runs locally with your own read-only credentials. Nothing is sent to CloudWise.

firstpass

The verification layer for LLM serving — your check runs on every answer before it ships, with a signed, tamper-evident receipt for each decision and a distribution-free bound on wrong answers served. Proof over prediction.