Skip to content

AI Agent Products

Key Points

Consumer / dev-facing AI products you might build against — or compete with. Buy-vs-build context.

  • ChatGPT (OpenAI), Claude.ai (Anthropic), Gemini (Google) — flagship chat assistants.
  • Perplexity — AI search.
  • Operator (OpenAI), Computer Use (Anthropic) — browser-driving agents.
  • Custom GPTs, Claude Projects, Gemini Gems — instance customizations.
  • Agent platforms: AutoGen (deprecated), CrewAI, LangGraph, Microsoft Agent Framework.
  • For senior devs: know what's possible to choose buy-vs-build.

Consumer chat assistants

ChatGPT

  • Free + Plus + Pro + Team + Enterprise tiers.
  • Custom GPTs (system-prompted variants + actions).
  • Voice mode, image gen (GPT-Image), web browsing.
  • Mac/Windows/Mobile native apps.

Claude.ai

  • Free + Pro + Team + Enterprise.
  • Projects (workspace with files + custom instructions).
  • Artifacts (rendered output).
  • Computer Use beta.

Gemini

  • Free + Advanced + Workspace integration.
  • Gems (custom personas).
  • Deep Research (long-form synthesis).
  • Multimodal native.

Perplexity

  • AI search engine.
  • Sources cited.
  • Pro tier with model choice.
  • API for embedding into apps.

Browser/computer agents

OpenAI Operator

Browser agent. User says "book a flight"; Operator browses, clicks, types. Limited rollout.

Anthropic Computer Use

Claude operates a virtual computer (screenshot + click + type). API-accessible.

For .NET dev: integration via Anthropic API.

Agent frameworks (build your own)

Microsoft Agent Framework

GA April 2026. .NET-native. Replaces Semantic Kernel + AutoGen for new work.

See Agent Framework section.

AutoGen

Older Microsoft Research project. Multi-agent. Superseded by Agent Framework for new work.

LangGraph (Python)

LangChain's stateful agent framework. Python-first; has interop.

CrewAI

Python; role-based multi-agent. Good for prototypes.

LlamaIndex

Python; RAG-focused; agents added.

For .NET 2026: Microsoft Agent Framework. Period.

Custom GPTs / Projects

For non-coding use: configure ChatGPT or Claude with system prompt + files. No code. Useful for internal "templates".

For coding/integration: build with Agent Framework or IChatClient + tools.

Buy vs build

Need Buy Build
Internal Q&A Custom GPT / Claude Project / Copilot Studio Custom RAG with Agent Framework
Customer-facing chat (custom build, or whitelabel) Agent Framework + RAG
Search Perplexity API or whitelabel RAG + reranker
Browser automation Operator / Computer Use API Custom with Playwright + LLM
Code agent Copilot / Claude Code / Cursor Custom CLI agent

Microsoft 365 Copilot

Embedded in Office apps. AI-assisted Word, Outlook, Teams, Excel, etc. Backed by GPT.

For organizations: licensing decision. For senior devs: Copilot Studio to extend with custom skills.

Copilot Studio

Low-code platform for building Copilot extensions, custom agents. Plugs into Microsoft 365 Copilot, Teams, websites. Bridges Power Platform + AI.

Salesforce Agentforce, Einstein

Salesforce's AI layer. Workflow automation in CRM context. Industry-specific.

Vertical agents

  • Cursor / Devin / Claude Code — coding.
  • Harvey — legal.
  • Hippocratic — healthcare.
  • Glean — enterprise search.

Many "AI X" products. Most either thin wrappers around frontier models, or fine-tuned domain models.

Buy-vs-build heuristic

Build if:
  - Customization is core to product.
  - Data is sensitive (sovereignty).
  - Need vendor portability.
  - Senior team available.

Buy if:
  - Generic use (Q&A, search).
  - Time-to-market critical.
  - Don't have AI expertise.
  - Budget for SaaS.

Senior consideration

Knowing the consumer landscape lets you: - Avoid building products already commoditized. - Position your custom solution against alternatives. - Pick partners (e.g., embed Perplexity vs roll your own search). - Spot rapid moves (frontier providers absorbing functions).

Cross-references