softwarefactory.build

A directory and guide library for agent-native software factories.

Directory / Claude Code

Claude Code

Verified

Anthropic's coding agent across terminal, IDE, desktop, and browser.

Claude Code is Anthropic's agentic coding tool: a CLI that reads a repository, edits files, runs commands, and opens pull requests, plus matching surfaces in VS Code, JetBrains, a desktop app, and the browser. Subagents, hooks, skills, CLAUDE.md memory, GitHub Actions, and scheduled routines make it a serious execution layer. It is not model-agnostic. Inference is Claude, wherever you host it.

Model independence
Locked
Deployment
Cloud, Self-hosted
Open source
No
License
Proprietary
Pricing model
Mixed
Added / updated
19 Aug 2026 / 19 Aug 2026

code.claude.com · Docs

Editorial take

Claude Code is the strongest single-harness execution tool in this seed set if your factory is allowed to standardize on Anthropic. The Unix posture — pipe logs in, run headless in CI, hook every file edit — is why teams build factories around it instead of an IDE. Locked models are the trade. Bedrock, Vertex, Foundry, and the Anthropic API change where Claude runs, not whether you can swap in another lab's model. Self-hosted environments exist in public beta on Team and Enterprise for cloud sessions; local CLI work already runs on the developer's machine. There is no public source repository for the product.

Strengths

  • One engine across terminal, IDE, desktop, and claude.ai/code, with CLAUDE.md, skills, hooks, and MCP shared between surfaces.
  • Headless and CI-friendly: GitHub Actions, GitLab CI, and the Agent SDK are documented first-class paths.
  • Self-hosted environments (public beta, Team/Enterprise) run cloud sessions on compute you operate; local sessions already run on your machine.
  • Verification hooks can block or format on every tool call, which is closer to a merge gate than a chat transcript.

Limitations

  • Model-locked. Third-party deployment options (Bedrock, Vertex, Foundry, AWS Marketplace) host Claude, they do not add GPT or Gemini as first-class models.
  • Not open source. The CLI is a distributed binary; there is no public repo to audit the harness.
  • No free product tier for Claude Code. Official docs require a Claude subscription or an Anthropic Console / third-party provider account. Public list prices on 19 August 2026 start at Pro $20/month; Max and Team Premium are substantially higher. Confirm on claude.com/pricing.
  • Self-hosted cloud environments are a beta, off by default, and exclude some surfaces. Do not plan an air-gapped factory on the CLI alone without reading the current limitations page.

Compare

  • vs cursor — The two default agent execution environments most teams actually argue about.
  • vs aider — Locked frontier harness versus the oldest model-agnostic CLI.
  • vs warp — Open terminal plus factories versus a locked, CI-friendly harness.
  • vs openai-codex — The two lab-shipped coding agents that lock you to a model family.

Fields last checked against primary sources on . Pricing and plan names rot; follow the vendor URL.