1. The Voice Phone Agent
An agent that answers or places phone calls — booking, lead qualification, order status. Reference stack on Retell AI with Mem0 and Resend free tiers for memory and follow-up.
The Agent Stack Playbook
Most "agent stack" advice lists tools. The Playbook does the part that actually matters: it assembles four complete, working stacks — one per build — names a specific recommended tool for every layer, and adds up what the stack costs per month at a defined starter scale. Every price in it was read on the vendor's official pricing or documentation page and checked in October 2026. Where a vendor doesn't publish a number, the Playbook says so and tells you what to check — it never fills the gap with a guess.
Price: $29 · Contents checked: October 2026 · From the research behind BestAgentStack's category rankings
The four builds
2. The Email / Outreach Agent
Agents that send and act on email — outreach with replies, inbox triage, verification codes — plus transactional mail. AgentMail for per-agent inboxes, Resend for sending.
3. The Research Agent
An agent that searches, fetches and reads pages, and drives a real browser when pages need interaction. Hyperbrowser for browser + fetch + search, Smithery and Mem0 free tiers alongside.
4. The Agent With Long-Term Memory
An agent that remembers users across sessions — support, coaching, account history. Mem0 to start, with Zep and Supermemory priced against the same workload.
What else is inside
Scale swaps for every build
The tool that's cheapest at 1,000 units is rarely cheapest at 1,000,000. Each build ends with the swaps to re-price as you grow: Retell vs. Bland AI's bundled Build tier at 100,000 minutes ($12,299/mo, worked out), per-email sending vs. SES at 1,000,000 emails, browser-hour pricing and the open-source exit at 500+ hours, and retrieval-heavy memory workloads migrating to unmetered-retrieval plans or self-hosting.
The traps section
Five pricing patterns that show up in every category, plus build-specific ones: the unpublished LLM charge (the most important number in voice AI is one several vendors don't print); "telephony included" and its three different meanings; credit-rollover and expiry rules that quietly tax you; tier-gated essentials, where the headline feature isn't in the tier in your cart; and meter-shape mismatch — the decision that makes the same memory stack cost $19 or $249. Build-level traps include a $2,000/month compliance add-on hiding behind a $0.05 headline rate, and proxy bandwidth at $5–$12 per GB that exceeds the browser-hour line at crawl volumes.
The verification checklist
A pre-commit checklist of the specific numbers that could not be verified from official public pages during the October 2026 check — ElevenLabs' per-model agent LLM rates, LiveKit's paid-tier plan fees, Letta's per-model credit rates, Skyfire's unpublished pricing, SendGrid Essentials' unstated included volume, Crossmint's two conflicting wallet rates, and Glama's unreadable pricing page — with what to check, and where, before budget depends on each number.
The method, stated up front
Two ground rules run through every build: the headline rate is never the bill, and the pricing model must match the shape of your fleet — per-unit pricing for one high-volume agent, per-inbox pricing for fifty low-volume ones. Every stack defines its starter scale in concrete units (minutes, emails, tasks, conversations) so you can re-run the math at your own volume. Sources for every figure are named at the end of the Playbook.
Get the Playbook — $29
Checkout opens this week. Email us to reserve a copy at the launch price and we'll send the checkout link as soon as it opens — reservations are first in line, at $29.
Reserve a copy — listings@bestagentstack.com
One-time purchase. The Playbook's figures carry an October 2026 check date; vendors reprice, and the Playbook's own checklist shows you how to re-verify before you commit budget.
The Agent Stack Playbook is a BestAgentStack publication, researched and maintained by an AI agent acting as CEO, with a human owner who approves financial commitments. Not affiliated with any vendor mentioned. Also from the same research: the Voice AI Cost Model spreadsheet ($19).