Claude Code

Claude Code is Anthropic's agentic coding tool. Its documentation describes it as a tool that "reads your codebase, edits files, runs commands, and integrates with your development tools", available in the terminal, in IDEs, in a desktop app and in the browser 1. You describe a task in plain language and Claude Code plans the approach, changes code across files and checks the result.

Where Claude Code runs

The same engine sits behind several surfaces 1:

• Terminal: the full-featured CLI, started with the claude command inside a project folder
• IDE: extensions for VS Code (and Cursor) and a plugin for JetBrains IDEs, with inline diffs
• Desktop app: a standalone app for running sessions side by side and reviewing diffs visually
• Web: sessions in the browser at claude.ai/code, with no local setup, useful for long-running tasks

Project instructions, settings and connected MCP servers carry across these surfaces 1.

What it can do

According to Anthropic's docs, typical work includes writing and running tests, fixing lint errors, resolving merge conflicts, tracing bugs to a root cause, and working with git: staging changes, writing commit messages, creating branches and opening pull requests 1. It can also run non-interactively with claude -p, so it can be piped into scripts or used in CI 1.

Key features

• MCP connections: Claude Code connects to external tools and data through the Model Context Protocol. Remote servers are added with claude mcp add --transport http at local, project or user scope, and servers that use OAuth are authenticated with the /mcp command 2.
• CLAUDE.md: a Markdown file of project instructions that Claude Code reads at the start of every session. It can also read an existing AGENTS.md 3.
• Subagents: specialised assistants that each run in their own context window with a custom system prompt, specific tool access and independent permissions 4.
• Skills: packaged, reusable workflows a team can share and invoke as commands 1.
• Hooks: shell commands that run before or after Claude Code actions, for example formatting after every edit 1.

Using Claude Code to draft and publish social posts

Because Claude Code can use any MCP server, it can work with social publishing tools as well as code. Publora, a social media scheduler with a publishing API, documents a Claude Code setup for its MCP server 5:

claude mcp add --transport http --scope user publora https://mcp.publora.com/mcp

After signing in, the server exposes 18 tools across 10 platforms (LinkedIn, X, Facebook, Instagram, Threads, TikTok, YouTube, Bluesky, Mastodon and Telegram), so Claude Code can list connected accounts, create or update posts, schedule them and read post statistics 5. API and MCP access are included on Publora's free plan 6.

A practical pattern is to keep campaign notes, brand rules and post drafts in a repository, ask Claude Code to turn a release note or blog post into platform-specific versions, save them as drafts, and have a person review them before they are scheduled. That keeps a human in the loop for anything that goes out publicly.

Safety and permissions

Anthropic advises verifying that you trust each MCP server before connecting it, because servers that fetch external content can expose you to prompt injection 2. Instruction files such as CLAUDE.md are treated as context, not enforced configuration; to block an action regardless of what the model decides, the docs recommend a hook instead 3. For social accounts, the safest setup combines drafts-first instructions with the publishing tool's own account permissions.