Prompt Caching Python Packages

agt-agent

An AI agent framework that builds itself — multi-model ReAct + tiered cache-friendly context + async multi-agent + XML workflows. CLI & WebUI, zero LangChain.

ccdrift

A check for silent changes in Claude Code's prompt caching, Haiku use, settings, context and hooks, read from your local session logs

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.

omnicache-proxy

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

replay-doctor

If the agent bill went up and nothing errored, a prompt cache broke. Replay Doctor names the turn it broke on, the cause, and the tokens re-billed at write prices. Reads Claude Code and Codex transcripts already on disk. Local Go CLI, zero third-party dependencies, no account, and no network request you did not type.

context-audit

Zero-config CLI that audits coding-agent context windows — measures cache health, repeated context, and structural waste from your local Claude Code / Cursor session logs. 100% local, no telemetry.

claudeglass

Are you using Claude Code well? Finds the habits and settings that waste tokens, from the transcripts on your machine. Local dashboard, stdlib-only Python.

sqz

Context compression for AI coding agents: compresses tool output before it enters the model, dedups repeats to 13-token refs. Claude Code, Cursor, Codex, Kiro, Zed, any MCP client. Rust, zero LLM calls.

cachecanary

Know the moment your Claude prompt cache breaks on Amazon Bedrock, not when the bill arrives.

codex-usage-tracking

Local-first MCP tools and dashboard for investigating Codex token usage, credits, costs, caching, and thread patterns.

pcdlint

Static Taint Linter for LLM Prompt-Cache Determinism (OpenAI & Anthropic)

tokenbill

Why is your agent bill so high? Profile LLM agent traces: token waterfalls, re-sent-prefix waste, prompt-cache simulation, and cache-breaker detection with dollar-valued fixes. Pure stdlib; the demo needs no API keys.

burnless

Stop replaying the transcript: capsule session state + rolling memory for LLM CLIs (Claude Code, Codex). Measured: −90.3% vs no-cache / −30% vs cached replay at turn 10, reproducible bench. MIT, provider-agnostic.

toolhunch

Your agent has too many tools. Hybrid search + a fast decision model (Jev, Laya) or an LLM pick the right one, or say "none". Pydantic AI integration and a ToolRet benchmark.

ctxbudgeter

ContextOps toolkit for production AI agents. Compile, govern, scan, visualize & optimize LLM context before every model call — token budgets, policy governance, PII/secret scanning, Context Bill of Materials, diffing, Context MRI, MCP budgeting. Framework-agnostic. Local-first. Deterministic.

cachemiss

Why did my Claude Code quota drain? Every prompt-cache rebuild in your local transcripts, with its reason, cost and culprit. No network, no dependencies.

cognisn-spark

Secure Personal AI Research Kit — Multi-provider LLM web interface with MCP tools, voice chat, persistent memory, and autonomous actions

convene

Run Claude Code headlessly as a local inference layer: named experts, multi-turn sessions, measured concurrency. Local and first-party only: no proxy, no network endpoint.

modelmeld

OpenAI- and Anthropic-compatible AI gateway. Capability-based routing. Streaming. BYOK passthrough. No key custody.

promptcachelint

Explain why your LLM prompt cache missed: prefix diffs, silent invalidators and hit-rate accounting for Anthropic and OpenAI.

justllm

Production LLM calls. Just the three lines. Cross-provider fallback, native caching, and reversible context compression on by default.

claude-cache-analyzer

CLI tool that measures prompt cache efficiency of Claude Code sessions

cachemux

One wrapper for LLM context caching across providers. Measures whether it pays, and sets it up when it does.

ctx-slim

Prune your LLM agent's context without destroying your prompt cache. Measured: 20.7% fewer tokens, 33-59% more money.