InkdownInkdown
Start writing

Yt Trans

15 files·15 subfolders

Shared Workspace

Yt Trans
0 Jobs To 4l Month The Ai Business That Changed My Life

tldr

Shared from "Yt Trans" on Inkdown

════════════════════════════════════════

MCP vs ACP: The Two Protocols Every AI Builder Needs to Know

Build Great Products · 35:09 · 2026-04-21

What This Is Actually About

MCP (Model Context Protocol) connects AI agents to tools like Notion and Slack. ACP (Agent Client Protocol) connects your application to CLI-based agents like Claude Code. They work together — not as alternatives — and together define the architecture for the next generation of agent-powered applications.

Key Points

MCP vs ACP: The Core Distinction

MCP is the bridge between agents and external tools/sources of context (Notion, Slack, Figma). ACP is the protocol that lets a client application (your frontend) talk directly to a CLI-based agent running in the terminal. MCP feeds the agent information; ACP gives the agent a UI.

The Terminal Problem ACP Solves

The terminal is the best place for agents to live — it gives them full access to the computer. But the terminal UI is an "awkward and horrible user experience" for most people. ACP lets you build a visual chat interface for terminal agents while the agent continues running in the CLI behind the scenes.

tldr.md
8 Lpa To 55 Lpa In 4 Months No Bs Breakdown Ft Dhairyasheel
tldr.md
Building The Perfect Linux Pc With Linus Torvalds
tldr.md
Claude Mythos Clone Shocks Anthropic And Openai
tldr.md
Don T Let Ai Rob You
tldr.md
English Or Spanish India S Got Latent
tldr.md
From Zero To Senior How I Grew In My Career
tldr.md
How to become 10x smarter
file
How To Learn Ai Engineering In 5 Minutes No Prior Knowledge
tldr.md
Mcp Vs Acp The Two Protocols Every Ai Builder Needs To Know
tldr.md
Optimize Your Ai Quantization Explained
tldr.md
Realistic Advice About Software Dev Right Now
tldr.md
Stop Using Ai For These Things
tldr.md
The Most Talented Man In Ai
tldr.md
What Is The Future Of Coding With Ai
tldr.md
The Architecture

User input → Client frontend → ACP → Agent (running as CLI subprocess) → MCP → Tools/Skills → Context flows back → Agent streams output to frontend via ACP. JetBrains and Google are the primary drivers of the protocol.

ACP-Compatible Agents

Claude Code, Codex, Gemini CLI, and Open Claw are already ACP-compatible. The ACP Agent Registry (from JetBrains) lets you browse and install agents directly from IDEs like JetBrains and Zed.

Building with ACP Requires Iteration

The speaker (self-described non-technical) built ACP into his app Orchestrate using Claude Code with Opus 4.6 in plan mode. The initial 15-minute plan had 6 phases. The build took ~30 minutes. Multiple bugs emerged — build failures, unhandled session update types, wrong package names for Codex, wrong flags for Gemini, MCP config mismatches. Each was fixed by feeding errors back to the agent.

ACP Enables a New Class of Desktop Apps

ACP lets any desktop app leverage local agents with the user's existing subscriptions. The speaker predicts AI-first reimaginings of legacy tools like Photoshop — using Claude Code on your computer to design on a canvas.

If You Remember Nothing Else

  • MCP connects agents to tools; ACP connects apps to agents.
  • ACP solves the UX problem of running agents directly in the terminal.
  • JetBrains and Google are driving ACP — it's not an Anthropic thing.
  • Expect bugs when integrating; feed errors back to the AI, don't restart from scratch.

════════════════════════════════════════