Weekly recap, June 1, 2026 – June 7, 2026

This week we published 48 production-grade AI tutorials — each a downloadable recipe that solves a real small-business problem with code you can run today. What follows is the full list: construction takeoffs, lien waiver collection, veterinary triage, and more. Everything is free to download and modify.

New tutorials

Bid Takeoff Agent for Small GCs

Estimators at small general contractors spend 2–3 days manually measuring plans and scoping subs. This agent ingests PDF plans and specs, extracts quantities via an LLM, and generates a structured bill of materials with subcontractor RFP drafts. It's built for GCs who want to cut bid prep from days to minutes.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag-pipeline, @reaatech/hybrid-rag-ingestion, @reaatech/agents-markdown, @reaatech/agent-eval-harness-golden, @reaatech/llm-router-core, @reaatech/context-window-planner on agnostic · 196 tests / 99.45% coverage.

Lien Waiver Collector for Specialty Trade Contractors

Project managers at specialty trades (electrical, plumbing, HVAC) lose hours each week chasing lien waivers before draws. This REST API automates requests, tracks status, and stores signed waivers with timestamps, using DocuSign for eSignatures and AI-drafted reminders.

Read the tutorial → Download the code (zip)

Built with @reaatech/a2a-reference-persistence, @reaatech/a2a-reference-auth, @reaatech/agent-memory-storage, @reaatech/idempotency-middleware, @reaatech/guardrail-chain, @reaatech/mcp-server-core on agnostic · 205 tests / 99.24% coverage.

Product Launch to Social Media Calendar

A B2B startup marketing lead spends two days a month crafting social posts from release notes. This Next.js app ingests product release notes (markdown with YAML frontmatter) and generates a 30-day social media calendar with platform-specific copy for LinkedIn, Twitter/X, and Facebook.

Read the tutorial → Download the code (zip)

Built with @reaatech/agents-markdown, @reaatech/agents-markdown-scaffold, @reaatech/agents-markdown-mcp-server, @reaatech/agents-markdown-validator, @reaatech/agents-markdown-reporter, @reaatech/agents-markdown-observability on agnostic · 85 tests / 98.55% coverage.

Support Ticket to Help Center Article Generator

An e-commerce support team with 200 weekly tickets can't keep up with writing help articles. This pipeline clusters closed Zendesk tickets by topic, extracts common Q&As, and generates draft articles ready for review using hybrid RAG and an LLM.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/hybrid-rag-ingestion, @reaatech/hybrid-rag-pipeline, @reaatech/hybrid-rag-retrieval, @reaatech/hybrid-rag-observability, @reaatech/hybrid-rag-mcp-server on agnostic · 173 tests / 99.34% coverage.

Agent Feedback Loop for Automated Fine-Tune Dataset Generation

A boutique marketing agency uses an AI agent for ad copy but manually edits outputs to match brand voice. This system logs agent outputs, captures user corrections (accept, reject, edit), and exports a clean dataset for fine-tuning a smaller, cheaper model.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-replay-core, @reaatech/agent-replay-interceptors, @reaatech/agent-replay-integrations, @reaatech/agent-replay-cli, @reaatech/agent-eval-harness-golden, @reaatech/agent-eval-harness-suite on agnostic · 113 tests / 98.18% coverage.

After-Hours Pet Emergency Triage Agent for Small Vet Practices

Vets dread after-hours calls because every limp or vomit demands immediate triage. This voice-enabled agent asks structured questions, classifies urgency, and generates a spoken recommendation, using speech-to-text, intent classification, and confidence-gated routing.

Read the tutorial → Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/voice-agent-telephony, @reaatech/voice-agent-stt, @reaatech/agent-mesh-classifier, @reaatech/agent-mesh-confidence, @reaatech/guardrail-chain-guardrails on agnostic · 74 tests / 99.24% coverage.

Auto-repair call intake agent for service advisors

Service advisors at independent auto-repair shops are constantly interrupted by price-estimate calls. This voice agent listens to the customer's description, identifies the repair type, provides a ballpark estimate, and escalates complex requests to a human.

Read the tutorial → Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/voice-agent-stt, @reaatech/voice-agent-tts, @reaatech/voice-agent-telephony, @reaatech/agent-handoff, @reaatech/agent-handoff-routing on agnostic · 119 tests / 100% coverage.

Per-tenant LLM cost tracking and chargeback for vertical SaaS

A B2B SaaS VP of Engineering needs to bill each SMB customer for AI features. This system attributes every LLM call to the right tenant, computes real-time spend, and exports cost data to billing, using OpenTelemetry and REAA cost packages.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-budget-spend-tracker, @reaatech/agent-budget-otel-bridge, @reaatech/llm-cost-telemetry, @reaatech/llm-cost-telemetry-exporters, @reaatech/otel-cost-exporter, @reaatech/multi-tenant-mcp-cost-accounting on agnostic · 74 tests / 100% coverage.

Anthropic Knowledge Agent for Shopify Product Q&A

Shopify store owners field repetitive product questions manually. This chat agent uses hybrid RAG over a synced product database to answer questions about specs, inventory, and orders, with multi-turn memory and session continuity.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/agent-memory, @reaatech/session-continuity, @reaatech/session-continuity-storage-dynamodb on anthropic · 111 tests / 99.64% coverage.

xAI Grok Observability for SMB AI Workflow Monitoring

SMBs using Grok lack visibility into token spend, latency, and errors. This Next.js app instruments every Grok API call with OpenTelemetry spans and cost telemetry, exporting data to Langfuse for a single-pane dashboard.

Read the tutorial → Download the code (zip)

Built with @reaatech/otel-genai-semconv-core, @reaatech/llm-cost-telemetry, @reaatech/otel-genai-semconv-instrumentation, @reaatech/llm-cost-telemetry-providers on xai-grok · 59 tests / 96.64% coverage.

Automated Industry Newsletter Curator

A consulting firm spends 5 hours weekly scanning RSS feeds for a client newsletter. This pipeline fetches articles, scores them by relevance, generates AI summaries, and compiles a draft — built with hybrid RAG and Hono.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/hybrid-rag-pipeline, @reaatech/hybrid-rag-retrieval, @reaatech/hybrid-rag-ingestion, @reaatech/hybrid-rag-observability, @reaatech/hybrid-rag-mcp-server on agnostic · 92 tests / 97.79% coverage.

xAI Grok Code Sandbox for Small Business Financial Modeling

Small business owners can't build complex financial models safely. This sandbox uses Grok to generate Python from plain English, then executes it in an isolated E2B sandbox with strict tool-use policies and cost tracking.

Read the tutorial → Download the code (zip)

Built with @reaatech/tool-use-firewall-core, @reaatech/llm-cost-telemetry, @reaatech/session-continuity, @reaatech/structured-repair-core, @reaatech/confidence-router on xai-grok · 97 tests / 97.74% coverage.

Mistral AI Document Pipeline for NetSuite Invoice Automation

Finance teams manually key invoice PDF data into NetSuite. This pipeline extracts text with OCR, parses line items with Mistral, repairs malformed JSON, and posts a draft vendor bill, with deduplication via agent memory.

Read the tutorial → Download the code (zip)

Built with @reaatech/media-pipeline-mcp-doc-extraction, @reaatech/structured-repair-core, @reaatech/agent-memory on mistral · 99 tests / 100% coverage.

Configurable PII redaction at ingress and egress for HIPAA, GDPR, and PCI compliance

A health-tech startup needs to prevent PHI leaks to LLMs. This dual-pipeline redacts PII before sending to the LLM and scans responses before showing to users, with switchable compliance profiles and audit logging.

Read the tutorial → Download the code (zip)

Built with @reaatech/guardrail-chain, @reaatech/guardrail-chain-guardrails, @reaatech/guardrail-chain-config, @reaatech/guardrail-chain-observability, @reaatech/mcp-gateway-validation, @reaatech/mcp-gateway-audit on agnostic · 125 tests / 98.59% coverage.

Review Response Agent for Independent Restaurants

Small restaurant owners manually reply to Yelp, Google, and TripAdvisor reviews, hurting response time and consistency. This agent generates on-brand replies using caching, guardrails, and agent memory.

Read the tutorial → Download the code (zip)

Built with @reaatech/a2a-reference-core, @reaatech/a2a-reference-client, @reaatech/agent-memory, @reaatech/llm-cache, @reaatech/guardrail-chain, @reaatech/a2a-reference-observability on agnostic · 108 tests / 95.37% coverage.

Automated Blog Content Pipeline with SEO Scoring

A SaaS founder spends 3 hours weekly copy-pasting drafts into a CMS and checking SEO. This API turns a topic brief into a validated, linted, and SEO-scored draft, with quality gates at each stage.

Read the tutorial → Download the code (zip)

Built with @reaatech/agents-markdown, @reaatech/agents-markdown-validator, @reaatech/agents-markdown-reporter, @reaatech/agents-markdown-linter, @reaatech/agents-markdown-mcp-server, @reaatech/agents-markdown-observability on agnostic · 89 tests / 100% coverage.

Slack/email triage to priority inbox

A 20-person support team gets buried in Slack DMs and emails. This agent reads all inbound messages, tags P0/P1/P2 urgency, and drafts AI replies for high-priority items, using intent classification and confidence routing.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-mesh-classifier, @reaatech/agent-mesh-confidence, @reaatech/agent-handoff-routing, @reaatech/agent-budget-llm-router-plugin, @reaatech/agent-mesh-utils on agnostic · 86 tests / 99.58% coverage.

Persistent customer memory across support sessions

Customer-support bots treat every chat as new, making users re-explain. This agent stores per-customer sessions, enforces a token budget, and auto-compresses older messages, using session-continuity and Claude.

Read the tutorial → Download the code (zip)

Built with @reaatech/session-continuity, @reaatech/session-continuity-storage-redis, @reaatech/session-continuity-storage-dynamodb, @reaatech/session-continuity-tokenizers on anthropic · 89 tests / 100% coverage.

Automated API key rotation for LLM providers with zero downtime

Dev tools startups rotate API keys every 90 days, but manual rotation causes outages. This sidecar rotates keys across AWS Secrets Manager or Vercel Env, with zero-downtime validation and audit logging.

Read the tutorial → Download the code (zip)

Built with @reaatech/secret-rotation-core, @reaatech/secret-rotation-sidecar, @reaatech/secret-rotation-provider-aws, @reaatech/secret-rotation-provider-vercel, @reaatech/secret-rotation-observability, @reaatech/secret-rotation-types on agnostic · 125 tests / 97.5% coverage.

Intake Automation Agent for Solo Immigration Attorney

A solo attorney spends 30+ minutes per client on manual intake and conflict checks. This agent screens clients, checks past cases via hybrid RAG, generates a case summary, and scrubs PII — dropping intake to 5 minutes.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-mesh, @reaatech/hybrid-rag, @reaatech/agent-memory, @reaatech/guardrail-chain, @reaatech/agent-handoff, @reaatech/llm-cache on agnostic · 139 tests / 99.11% coverage.

Deposition Prep Summarizer for Plaintiff Litigation Paralegal

A paralegal spends unbillable hours reading deposition transcripts. This app ingests PDFs, chunks text, and uses an LLM with agent memory to produce citable summaries, with cost tracking and context-window planning.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/agent-eval-harness-suite, @reaatech/agent-memory, @reaatech/llm-cost-telemetry, @reaatech/context-window-planner, @reaatech/agent-replay on agnostic · 76 tests / 96.93% coverage.

Google Gemini Voice Agent for Clinic Appointment Scheduling

Small clinics miss after-hours calls and get overwhelmed during peaks. This voice AI receptionist answers calls via Twilio, checks availability, books appointments, and sends SMS reminders — all with Gemini, Deepgram, and Cartesia.

Read the tutorial → Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/voice-agent-mcp-client, @reaatech/voice-agent-telephony, @reaatech/session-continuity on google-gemini · 155 tests / 97.7% coverage.

Mistral AI Document Pipeline for Xero Expense Report Processing

Small businesses manually enter receipt data into Xero. This pipeline extracts line items from PDFs and Excel sheets using Mistral, repairs malformed JSON, enforces budgets, and pushes validated expenses as ACCREC invoices.

Read the tutorial → Download the code (zip)

Built with @reaatech/structured-repair-core, @reaatech/llm-cost-telemetry, @reaatech/context-window-planner, @reaatech/agent-budget-engine on mistral · 60 tests / 100% coverage.

Google Gemini Runbook Automation for PagerDuty SMB Incidents

Small DevOps teams rarely have written runbooks. This CLI pulls PagerDuty incident data, uses Gemini to generate narrative, and outputs markdown runbooks with duplicate detection and Inngest scheduling.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-runbook, @reaatech/agent-runbook-analyzer, @reaatech/agent-runbook-alerts, @reaatech/agent-runbook-health-checks, @reaatech/agents-markdown, @reaatech/confidence-router-core on google-gemini · 73 tests / 95.9% coverage.

OpenAI Lead Intake for BigCommerce Small Business Sales

A single staff member can't triage all contact-form inquiries in real time. This API classifies intent, routes leads, generates AI auto-replies, and notifies Slack, with session continuity and cost tracking.

Read the tutorial → Download the code (zip)

Built with @reaatech/confidence-router, @reaatech/agent-handoff, @reaatech/agent-handoff-routing, @reaatech/session-continuity, @reaatech/structured-repair-core on openai · 79 tests / 96.74% coverage.

Cohere Agent Mesh for Multi-Channel Customer Support Triage

Support teams juggle email, chat, and social media with manual routing. This agent mesh classifies tickets with Cohere, routes by confidence to specialist agents, and tracks per-channel LLM costs.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-mesh, @reaatech/agent-mesh-router, @reaatech/agent-mesh-confidence, @reaatech/confidence-router-core, @reaatech/llm-cost-telemetry on cohere · 99 tests / 95.04% coverage.

OpenAI Guardrail Layer for SMB Customer Chat Safety

Small business chatbots face prompt injection and PII leaks. This drop-in shim wraps the OpenAI SDK with configurable guardrails for injection, PII redaction, and toxic content filtering.

Read the tutorial → Download the code (zip)

Built with @reaatech/guardrail-chain, @reaatech/guardrail-chain-config, @reaatech/guardrail-chain-guardrails, @reaatech/guardrail-chain-observability on openai · 55 tests / 99.13% coverage.

OpenAI Voice Agent for Aircall Small Business Support

Small businesses lose calls after hours. This voice receptionist answers Aircall calls, handles FAQs, and escalates to humans, with speech-to-text and text-to-speech via Deepgram and OpenAI.

Read the tutorial → Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/voice-agent-stt, @reaatech/voice-agent-tts, @reaatech/agent-handoff, @reaatech/llm-cost-telemetry, @reaatech/structured-repair-core on openai · 70 tests / 97.4% coverage.

Anthropic Document Pipeline for Klaviyo SMB Email Campaigns

Marketers spend hours manually creating email campaigns. This pipeline loads product feeds from Klaviyo, drafts personalized content with Claude, repairs output, and pushes campaigns back to Klaviyo.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/structured-repair-core, @reaatech/agent-budget-engine, @reaatech/llm-cache, @reaatech/prompt-version-control, @reaatech/llm-cost-telemetry on anthropic · 92 tests / 99.55% coverage.

Anthropic LLM Observability for SMB AI Operations

SMBs using Claude lack visibility into spend and errors. This instrumentation layer wraps every Anthropic SDK call with OTel spans and cost telemetry, enforcing budgets and exporting to Langfuse or Phoenix.

Read the tutorial → Download the code (zip)

Built with @reaatech/otel-genai-semconv-anthropic, @reaatech/llm-cost-telemetry, @reaatech/otel-genai-semconv-core, @reaatech/otel-genai-semconv-exporters, @reaatech/agent-budget-otel-bridge on anthropic · 85 tests / 100% coverage.

Cohere MCP Server for SMB Research and Summarization

Small businesses need AI agents that can research and summarize. This MCP server exposes Cohere's language models and Tavily search as composable tools, with auth and observability.

Read the tutorial → Download the code (zip)

Built with @reaatech/mcp-server-core, @reaatech/mcp-server-engine, @reaatech/mcp-server-tools, @reaatech/mcp-server-auth, @reaatech/mcp-server-observability on cohere · 56 tests / 97.54% coverage.

Databricks Agent Eval Harness for SMB Support Bots

SMBs deploying support bots need automated regression testing. This CI-friendly harness evaluates agent responses against golden datasets, scores with Databricks LLM, enforces quality gates, and exports to Braintrust.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-eval-harness-golden, @reaatech/agent-eval-harness-judge, @reaatech/agent-eval-harness-cli, @reaatech/agent-eval-harness-cost, @reaatech/agent-eval-harness-gate on databricks · 69 tests / 97.39% coverage.

Return Reason Agent for Shopify Merchants

An ops manager spends 3+ hours daily manually reviewing return reasons. This agent automates refund, replacement, and RMA decisions, using guardrails, hybrid RAG, and structured repair to apply store policy instantly.

Read the tutorial → Download the code (zip)

Built with @reaatech/structured-repair-core, @reaatech/structured-repair-mcp, @reaatech/hybrid-rag-pipeline, @reaatech/hybrid-rag-ingestion, @reaatech/context-window-planner, @reaatech/guardrail-chain-guardrails on agnostic · 72 tests / 100% coverage.

Behavioral Email Trigger Agent for WooCommerce Merchants

A marketing contractor relies on fixed weekly emails, missing cart abandoners. This agent classifies browsing events in real time, stores behavioral memory, and triggers personalized emails via Klaviyo.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-memory, @reaatech/agent-memory-core, @reaatech/agent-memory-retrieval, @reaatech/webhook-relay-core, @reaatech/webhook-relay-tools on agnostic · 76 tests / 96.66% coverage.

Anthropic Knowledge Agent for PayPal SMB Transaction Insights

Owners waste time manually searching PayPal for answers like “largest refunds last month.” This Q&A agent ingests transaction history into Qdrant, uses hybrid RAG with Claude, and maintains multi-turn context.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/hybrid-rag-pipeline, @reaatech/hybrid-rag-qdrant, @reaatech/agent-memory, @reaatech/session-continuity, @reaatech/llm-cost-telemetry on anthropic · 108 tests / 99.63% coverage.

Azure AI Document Pipeline for Sage Intacct Invoice Automation

SMBs manually re-key PDF invoices into Sage Intacct. This pipeline extracts text, parses fields with Azure OpenAI, repairs JSON, and posts AR entries, flagging low-confidence fields for review.

Read the tutorial → Download the code (zip)

Built with @reaatech/structured-repair-core, @reaatech/confidence-router-core, @reaatech/llm-cache, @reaatech/llm-cost-telemetry, @reaatech/media-pipeline-mcp-doc-extraction on azure-ai · 99 tests / 97.64% coverage.

xAI Grok Agent Eval Harness for SMB Support QA

Small businesses using Grok for support agents lack automated QA. This eval harness runs batch tests against golden datasets, scores with Grok, enforces quality gates in CI, and streams traces to Langfuse.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-eval-harness-suite, @reaatech/agent-eval-harness-judge, @reaatech/agent-eval-harness-gate, @reaatech/agent-eval-harness-observability on xai-grok · 91 tests / 100% coverage.

LangChain Reliability Suite for SMB Support Agents with Auto-Runbooks

LLM outages and rate limits can take support agents offline. This suite adds circuit breakers, idempotent workflows, auto key rotation, and self-generating runbooks, keeping agents up 24/7.

Read the tutorial → Download the code (zip)

Built with @reaatech/circuit-breaker-agents, @reaatech/idempotency-middleware, @reaatech/session-continuity-storage-dynamodb, @reaatech/agent-runbook, @reaatech/secret-rotation-core on langchain · 90 tests / 94.38% coverage.

Listing Copy Multiplier for Residential Brokers

A solo broker spends 45+ minutes rewriting listings for MLS, Zillow, social, and print. This API generates four platform-optimized variants from one draft, using hybrid RAG and caching for under-2-minute turnaround.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/agents-markdown, @reaatech/llm-cache, @reaatech/agent-eval-harness-golden, @reaatech/llm-router-core, @reaatech/otel-genai-semconv-core on agnostic · 156 tests / 98.08% coverage.

Anthropic Document Pipeline for Sage Intacct SMB Invoice Processing

A second pipeline for Sage Intacct: this one uses Claude to extract line items from PDFs, repairs output, and routes by confidence to auto-post or human review, with budget enforcement.

Read the tutorial → Download the code (zip)

Built with @reaatech/media-pipeline-mcp-doc-extraction, @reaatech/structured-repair-core, @reaatech/confidence-router, @reaatech/agent-budget-engine, @reaatech/llm-cost-telemetry on anthropic · 73 tests / 98.05% coverage.

OpenRouter Voice Agent for Cal.com Appointment Scheduling

Small businesses lose appointments when customers can't book over the phone. This voice agent lets callers book, reschedule, or cancel Cal.com appointments via phone, with OpenRouter for provider flexibility.

Read the tutorial → Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/voice-agent-stt, @reaatech/voice-agent-tts, @reaatech/voice-agent-telephony, @reaatech/confidence-router, @reaatech/agent-budget-engine on openrouter · 155 tests / 98.92% coverage.

Google Gemini Multi-Agent Handoff for Gorgias Support Triage

Support teams using Gorgias get buried when ticket volume spikes. This multi-agent system classifies intent, answers FAQs with Gemini, and escalates complex issues to a human with full context compression.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-handoff, @reaatech/agent-handoff-compression, @reaatech/agent-handoff-protocol, @reaatech/agent-handoff-routing, @reaatech/agent-mesh-classifier, @reaatech/agent-mesh-observability on google-gemini · 73 tests / 98.86% coverage.

OpenAI Cost Control for SMB Agent Workflows

OpenAI bills climb from repeated questions. This system wraps every call with cost telemetry, enforces monthly budgets with hard stops, and caches semantically similar responses using Redis.

Read the tutorial → Download the code (zip)

Built with @reaatech/llm-cost-telemetry, @reaatech/agent-budget-engine, @reaatech/llm-cache, @reaatech/llm-cache-adapters-redis, @reaatech/llm-cost-telemetry-aggregation on openai · 84 tests / 99.46% coverage.

Cohere Document Pipeline for HR Policy Compliance

Small businesses manually cross-reference policy PDFs against regulations. This pipeline ingests documents, extracts structured clauses with Cohere, repairs JSON, and surfaces compliance gaps.

Read the tutorial → Download the code (zip)

Built with @reaatech/media-pipeline-mcp-doc-extraction, @reaatech/media-pipeline-mcp-core, @reaatech/structured-repair-core on cohere · 90 tests / 92.09% coverage.

Cohere Agent Mesh for Insurance Quote Comparison

Independent agents manually enter client data into each carrier portal. This multi-agent system queries carriers in parallel, compares coverages, and recommends the best policy, orchestrated by Cohere.

Read the tutorial → Download the code (zip)

Built with @reaatech/agent-mesh, @reaatech/agent-mesh-router, @reaatech/agent-mesh-registry, @reaatech/agent-mesh-session, @reaatech/agent-mesh-observability on cohere · 106 tests / 93.82% coverage.

AWS Bedrock MCP Server for GitHub Small Business DevOps

Small dev teams waste time navigating GitHub. This MCP server connects Bedrock agents to GitHub, exposing safe read tools with a destructive-operation firewall.

Read the tutorial → Download the code (zip)

Built with @reaatech/mcp-server-core, @reaatech/mcp-server-tools, @reaatech/mcp-server-auth, @reaatech/tool-use-firewall-core on bedrock · 103 tests / 96.17% coverage.

Cohere Knowledge Agent for Plaid SMB Financial Insights

Owners manually comb statements for cash flow answers. This Q&A agent connects Plaid-linked accounts, indexes transactions into Qdrant, and answers natural-language questions with Cohere RAG.

Read the tutorial → Download the code (zip)

Built with @reaatech/hybrid-rag, @reaatech/session-continuity, @reaatech/structured-repair-core, @reaatech/confidence-router on cohere · 135 tests / 98.38% coverage.

Databricks Code Sandbox for Secure SMB Data Analysis

Non‑technical staff can't query data safely. This sandbox translates plain English into SQL/Python, dry-runs in E2B, enforces security policies, and executes on Databricks, with cost tracking.

Read the tutorial → Download the code (zip)

Built with @reaatech/confidence-router, @reaatech/structured-repair-core, @reaatech/tool-use-firewall-core, @reaatech/llm-cost-telemetry, @reaatech/session-continuity on databricks · 103 tests / 96.49% coverage.

Browse all 48 solutions →

Building blocks shipped

Fifty-four npm package versions shipped across the week, mostly security patches, dependency upgrades, and reliability fixes. Here's what changed, grouped by repo.

Agent Budget Controller

The budget enforcement stack (@reaatech/agent-budget-engine, agent-budget-pricing, agent-budget-types) got stability fixes: a null-guard for minimal policies, cache-token pricing support, and a zod v4 upgrade. All three patches are additive and backwards-compatible.

View the Agent Budget Controller family →

Agents MD Kit

All nine packages in the markdown toolchain for agents — from core types to the MCP server — received patches to resolve a CI security audit failure, restoring green builds. agents-markdown, agents-markdown-validator, and agents-markdown-mcp-server are the most commonly used.

View the Agents MD Kit family →

Circuit Breaker Agents

@reaatech/circuit-breaker-core@0.1.1 fixes a bug where failureStrategy was silently ignored and improves error handling for the common misuse case.

View the Circuit Breaker Agents family →

Classifier Evals

Eight packages that form the classifier evaluation framework — types, CLI, judge, metrics, gates, and exporters — were updated to fix a CI security audit by upgrading vitest and related dev tooling. No runtime changes for consumers.

View the Classifier Evals family →

Hybrid RAG Qdrant

hybrid-rag-mcp-server migrated to zod 4, and hybrid-rag-observability upgraded its OpenTelemetry SDKs to 2.x, improving compatibility.

View the Hybrid RAG Qdrant family →

Idempotency Middleware

@reaatech/idempotency-middleware-adapter-firestore@1.1.0 bumped its Firestore runtime dependency to v8, a major version change that may affect existing consumers using v7.

View the Idempotency Middleware family →

LLM Cost Telemetry

@reaatech/llm-cost-telemetry upgraded zod to v4, requiring explicit type arguments for z.record in its public schema. The CLI tool also bumped Commander to v14. Both are minor updates.

View the LLM Cost Telemetry family →

LLM Judge Toolkit

Ten judging packages — from bias detectors to calibration, consensus, and providers — picked up a security-audit fix, plus various dependency upgrades including zod v4 in llm-judge-types and llm-judge-calibration, and pino v10 in llm-judge-infra.

View the LLM Judge Toolkit family →

LLM Router

@reaatech/llm-router-engine@1.0.1 fixes a corrupted package.json dependency line that could break installs. Patch release only.

View the LLM Router family →

MCP Contract Kit

All six contract validation packages — core, validators, reporters, client, CLI, and observability — addressed a critical CVE in vitest and received a minor commander bump in the CLI.

View the MCP Contract Kit family →

MCP Server Starter TS

@reaatech/mcp-server-core@1.0.1 moved to zod v4 for security, while mcp-server-observability@1.1.0 upgraded OpenTelemetry APIs to 2.x, requiring migration from the deprecated Resource constructor.

View the MCP Server Starter TS family →

Media Pipeline MCP

Three provider packages saw major dependency shifts: media-pipeline-mcp-deepgram moved to Deepgram SDK v5, media-pipeline-mcp-google bumped Vertex AI and Document AI SDKs by multiple majors, and media-pipeline-mcp-replicate upgraded the Replicate SDK to v1.4. Consumers should test carefully.

View the Media Pipeline MCP family →

Webhook Relay MCP

Five packages debuted on npm this week: webhook-relay-core, webhook-relay-mcp, webhook-relay-storage, webhook-relay-tools, and webhook-relay-webhooks. The monorepo restructure brought production-grade features like API key auth, SSE transport, delivery retry, and source heartbeat monitoring.

View the Webhook Relay MCP family →

Browse the full catalog at reaatech.com/products.