AgentHubAgentHub

AI tools·Coding agents·Global

Claude Code

A terminal coding agent: read the repo, edit, test. Tightly bound to Claude models and MCP.

9.6/ 10.0Editorial Rating
Editorial score based on public benchmarks and community feedback — not a third-party benchmark result

Compiled through 2026-09

Claude Code visual
Overseas accessPaidCodingAutonomous agentCompare
Vendor
Anthropic
Started
2025
China availability
Overseas accessRequires overseas network access
Pricing tier
Paid
Current flagship
Claude Code + Claude 3.7 Sonnet / Fable 5.1
Website
docs.anthropic.com/en/docs/claude-code

Evaluation, Pros & Cons

Editorial score based on public benchmarks and community feedback — not a third-party benchmark result

Key Advantages

  • Anthropic official CLI agent operating directly in terminal shell to edit and test
  • Deep synergy with Claude 3.7 hybrid reasoning and MCP; unmatched CLI autonomy

Limitations & Caveats

  • Terminal-only interface; requires comfort with CLI workflows

Overview

Claude Code is Anthropic’s CLI coding agent: plan, tool-call, MCP, plus CLAUDE.md and Skills. Unlike Cursor it is terminal-first, not a full IDE. Opus 5 and Fable 5.1 in 2026 made long refactors more reliable. AgentHub publishes Claude Code install snippets for Skills and MCP.

Direction

It will keep riding Claude flagships (computer use, long context, adaptive thinking) and the MCP/Skill ecosystem. Same model family as claude.ai and the API.

Version history

  1. Sep 2026

    Latest

    Fable 5.1

    1M context default for huge repos.

    View details

    New capabilities

    • Fable 5.1 with 1M context by default

    What got stronger

    • Less “summarize the tree first” preprocessing on huge monorepos

    Version notes

    5.1’s meaning for Claude Code is blunt: default context is large, adaptive thinking picks duration. Quotas and price stay official.

    #2026-09-fable-5-1

  2. Jul 2026

    Opus 5 / Fable 5

    Long-horizon coding as the main job.

    View details

    New capabilities

    • Opus 5 / Fable 5 for long-horizon coding

    What got stronger

    • “Refactor this subsystem” became one job, not a pile of chats

    Version notes

    The Fable line exists for Code-style leave-it-running jobs. After July, long tasks were the story.

    #2026-07-opus-5-fable-5

  3. Sep 2025

    Sonnet 4.5 agents

    More reliable repo tasks.

    View details

    New capabilities

    • Sonnet 4.5 as the default workhorse

    What got stronger

    • Fewer blow-ups on real repo tasks

    Version notes

    4.5 made many teams daily-drive Claude Code. Opus stayed for the hardest architecture jobs.

    #2025-09-sonnet-4-5-agents

  4. May 2025

    Claude 4 pairing

    Jumps with Opus/Sonnet 4.

    View details

    New capabilities

    • Paired with Claude 4 tool loops

    What got stronger

    • Same terminal, stronger model, visibly higher success

    Version notes

    Claude 4 turned Code from a demo into something that could edit a mid-size repo. MCP and permissions started to matter.

    #2025-05-claude-4-pairing

  5. Feb 2025

    Claude Code preview

    Terminal coding agent preview.

    View details

    New capabilities

    • Read the repo, edit files, run commands in a terminal
    • CLAUDE.md project notes

    What got stronger

    • Agents without switching IDE

    Version notes

    The preview made “open Claude in this directory” a workflow. Good for tmux/CLI people and for calling it from CI.

    #2025-02-claude-code-preview