Appearance
Chapter 7 — Toolkit Agent 2026
🧰
"40+ tools, 6 layers. Pick the layer first, the tool second."
How to use this chapter
- Read overview tables — know the landscape
- Quick picker — fast tool choice for use case
- Re-check every 3 months — Anthropic + OpenAI release fast
01 Layer 1: Coding Agents
| Tool | Price (May 2026) | Strength | Best for |
|---|---|---|---|
| Claude Code | $20 Pro / $100-200 Max / API | Terminal-native, sub-agent, 30h autonomous, MCP-native | Multi-file refactor, large codebase |
| Cursor | $20/month Pro (credit-based) | Best autocomplete, polished IDE | Daily inline coding |
| Windsurf (Cascade) | $15/month Pro (cheapest top 3) | "Flows" persistent context | Value-conscious |
| Devin 2.0 | $20 Core / $500 Team | Cloud-based, parallel autonomous PR | Long tickets, enterprise |
| OpenAI Codex | Bundle ChatGPT Plus/Pro | 2M+ WAU Mar 2026, cloud sandbox | ChatGPT users, cloud async |
| GitHub Copilot Workspace | $10-39/user/month | GitHub-native | Existing GH enterprises |
| Replit Agent | $25/month Core | Browser-based, full-stack scaffold | Non-dev MVP |
| Aider | Free + your API key | 45.3K stars, terminal, git-aware | OSS purists |
| Cline | Free + your API key | 62.3K stars, "trusted by 8M+", VS Code | Visual diff approval |
Quick picker
| Use case | Top pick | Backup |
|---|---|---|
| Multi-file refactor large codebase | Claude Code | Devin |
| Daily inline coding flow | Cursor | Windsurf |
| Budget-tight | Windsurf or Aider | Cline |
| Cloud async (long task) | Devin or OpenAI Codex | — |
| Non-dev build MVP | Replit Agent | Lovable |
| OSS / privacy concern | Aider or Cline | — |
02 Layer 2: General-Purpose Agents
| Tool | Price | Status (May 2026) |
|---|---|---|
| Manus AI | $199/month Pro | Blocked by Chinese NDRC Apr 2026 — Meta acquisition $2-3B unresolved |
| Anthropic Computer Use | API Sonnet 4.6 $3/$15 | Beta computer-use-2025-11-24; 72.5% OSWorld-Verified |
| OpenAI ChatGPT Workspace Agents | Bundle enterprise | Apr 2026 — Operator absorbed |
| Project Mariner | — | Shutdown May 4, 2026 — absorbed into Gemini Agent |
| Claude Code SDK | API-priced | Orchestrator-via-Task tool, MCP-native |
Quick picker
| Use case | Top pick |
|---|---|
| Automate legacy app no-API | Anthropic Computer Use |
| ChatGPT-native workflow | Workspace Agents |
| Build custom agent | Claude Code SDK |
03 Layer 3: Workflow / Automation
| Tool | Price | Numbers (May 2026) | Best for |
|---|---|---|---|
| n8n | Free / $20+ cloud / $13.3K avg ARR enterprise | $40M ARR, $180M Series B, $2.5B valuation, 230K users, 3K+ enterprise | Full-stack workflow, AI nodes |
| Make | $9-29/month | 1,500+ apps | Visual scenarios |
| Zapier Central/Agents | Variable | Established | Existing Zapier users |
| Smax.ai (Vietnam) | $25-89/month | Yody +15-20% close, Let's Sushi +300% online | VN Messenger/Zalo bot |
| Lindy | Free / $49.99 Pro / $299.99 Business | 2,300+ apps | No-code agent |
| Sema4.ai | Enterprise | $55.5M raised, Emerson + Koch | Finance/HR enterprise |
Quick picker
| Use case | Top pick |
|---|---|
| Full-stack workflow + AI | n8n |
| Vietnamese Messenger/Zalo | Smax.ai |
| No-code agent for all | Lindy |
| Visual scenarios non-tech | Make |
| Enterprise SAFE | Sema4.ai |
04 Layer 4: Multi-Agent Frameworks
| Framework | Status May 2026 | Best for |
|---|---|---|
| LangGraph | Production powerhouse — Klarna, Uber, LinkedIn. Surpassed CrewAI stars early 2026. | Mission-critical production |
| CrewAI | Role-based DSL, 20 lines to start. 60% Fortune 500, 44K+ stars | Prototype, agency |
| AutoGen 0.4+ | Microsoft maintenance mode | Don't start new |
| OpenAI Swarm → Agents SDK | Lightweight handoff-based | OpenAI-native |
| Anthropic / Claude Code SDK | Orchestrator-worker via Task tool, MCP-native | Claude-first stack |
| A2A protocol | Open standard Apr 2025, donated Linux Foundation Jun 2025, 150+ supporters | Cross-vendor agents |
Quick picker
| Use case | Top pick |
|---|---|
| Production scale | LangGraph |
| Prototype fast | CrewAI |
| Claude-native | Claude Code SDK |
| OpenAI-native | OpenAI Agents SDK |
| Cross-vendor agents | A2A |
05 Layer 5: Infrastructure
| Tool | Differentiator | Customers |
|---|---|---|
| E2B | microVM, ~150ms cold start, 500M+ sandboxes processed, 88% Fortune 100 signed | Perplexity, Hugging Face, Groq, Manus |
| Browserbase | Headless browser as a service, 1K+ paying customers in 16 months, $40M Series B at $300M valuation | — |
| Daytona | Docker-based, <90ms sandbox creation, $24M Series A Feb 2026 | — |
| Modal / Vercel Sandbox | Serverless compute for agents | — |
| MCP servers ecosystem | 5,800+ official, 17,468 cross-registry, 97M monthly SDK downloads | Anthropic, OpenAI, Google all adopted |
Quick picker
| Use case | Top pick |
|---|---|
| Code execution sandbox | E2B |
| Browser automation | Browserbase |
| Cheap Docker sandbox | Daytona |
| MCP server | Anthropic SDK + Smithery |
06 Layer 6: Voice Agents
| Tool | Latency | Cost/min | Strength |
|---|---|---|---|
| ElevenLabs Conversational AI | <100ms | $0.08-0.24 | 11K+ voices, 70+ languages |
| Vapi | ~150ms | $0.05 orch + provider | Provider-agnostic, 62M monthly calls, 99.99% SLA |
| Retell AI | ~620ms | $0.07+ flat | HIPAA included, no-code + SDK |
| Bland | ~400ms | Variable | Phone-first |
Quick picker
| Use case | Top pick |
|---|---|
| Multi-language brand | ElevenLabs |
| Custom stack + scale | Vapi |
| Healthcare HIPAA | Retell |
| Outbound campaigns | Bland |
07 5 ideal stacks by persona
Persona 1: Solo dev (budget < $100/month)
Coding: Cursor Pro ($20) or Claude Code Pro ($20)
LLM API: Claude Haiku 4.5 + Sonnet 4.6 (~$30/month usage)
Sandbox: E2B free tier
Deploy: Vercel free + Supabase free
─────────────
Total: ~$60-80/monthPersona 2: Indie agent builder
Coding: Claude Code Max ($100)
Workflow: n8n cloud Starter ($20)
Infra: E2B ($20) + Browserbase free tier
Voice: Vapi ($30 usage)
─────────────
Total: ~$170/monthPersona 3: Agency (serving SMEs)
Coding: Claude Code Pro ($20)
Workflow: Smax.ai Pro ($89) + n8n Pro ($50)
LLM: Claude API ($50-200 usage)
Voice: Vapi ($50-200)
CRM: Pancake ($30)
─────────────
Total: ~$280-580/month (charge $5-15K project)Persona 4: Enterprise team (10+ devs)
Coding: Claude Code Max × 10 devs ($1,000-2,000)
Workflow: n8n Enterprise ($1K+)
Multi-agent: LangGraph (LangSmith $250+)
Voice: Vapi enterprise ($500+)
Eval: Braintrust ($500)
─────────────
Total: ~$3K-5K/monthPersona 5: Operator non-dev
Workflow: Lindy Pro ($49.99) or Make ($29)
LLM: ChatGPT Plus ($20)
Voice: Retell no-code tier
─────────────
Total: ~$70-100/month08 Selection framework
Decision tree
Q1: What's the use case?
- Solo coding → Layer 1 (Claude Code / Cursor)
- Automate workflow → Layer 3 (n8n / Smax.ai)
- Custom agent product → Layer 4-5 (LangGraph + E2B)
- Voice → Layer 6 (Vapi / ElevenLabs)
Q2: Budget?
- < $100/month → free tier + 1 Pro tool
- $100-500/month → multi-layer indie stack
$500/month → enterprise tier or team
Q3: Regional context needed?
- Yes (target VN customer) → Smax.ai + Zalo OA + Pancake mandatory
- No (global) → standard global stack
Q4: Skill level?
- Non-dev → Lindy + Make + Replit Agent
- Junior dev → Cursor + n8n + ChatGPT
- Senior dev → Claude Code + LangGraph + custom MCP
09 Continue reading
- Chapter 1 — Vibe Coding Solo
- Chapter 2 — Claude Code Deep
- Chapter 3 — Computer Use
- Chapter 4 — Multi-Agent
- Chapter 5 — Workflow Agent
- Chapter 6 — MCP
- Chapter 8 — Safety & Evals
- Chapter 9 — 30-Day Roadmap
Final word
"Don't memorize tools. Memorize layer roles.Coding agent = augment human dev.Computer Use = automate legacy.Workflow = orchestrate cross-app.Multi-agent = parallel specialists.MCP = standard interconnect.Voice = phone/call interface.Tools rotate. Layers stay."