Skip to content
Documentation
Cookbook
Integrations & Ops: Third Party, Fine-Tuning, and Cost Tracking

Integrations & Ops: Third Party, Fine-Tuning, and Cost Tracking

These cookbooks focus on putting Claude into real systems: third-party integrations, cloud platform workflows (e.g., Bedrock), and usage/cost observability.

Overview

Building production applications with Claude requires more than just API calls:

  • Integrations: Connect Claude with databases (Pinecone, MongoDB), search services (Wikipedia), and specialized tools (Wolfram Alpha, Deepgram)
  • Cloud Platforms: Deploy on AWS Bedrock with fine-tuning capabilities
  • Operations: Track usage, monitor costs, and set up observability for production workloads

Third-party integrations

Pinecone RAG
RAG using Pinecone
MongoDB RAG
RAG using MongoDB
LlamaIndex basic RAG
Basic RAG pipeline with LlamaIndex
Wikipedia search
Iterative searching Wikipedia with Claude
Wolfram Alpha tool
Use WolframAlpha as a tool
ElevenLabs voice assistant
Low-latency voice assistant
Deepgram transcription
Audio transcription + question generation

Fine-tuning

Finetuning on Bedrock
Fine-tune Claude 3 Haiku on Amazon Bedrock
Fine-tuning examples often involve sensitive datasets. Align on compliance, redaction, permissions, and retention before uploading data.

Usage & cost observability

Usage & cost Admin API
Usage/cost tracking scaffolding

This notebook is a good starting point for a minimal cost-tracking scaffold (grouping/filtering/pagination/export), before wiring into your own monitoring and alerting stack.