Claude Code CLI as an Agent Platform

Claude Code isn't just a coding assistant — it's a general agent runtime with MCP, subagents, hooks, and skills. Treat it that way and you get a free, powerful platform.

48 min · Reviewed 2026

More than a coding tool

Claude Code ships as a developer CLI, but underneath it's a general agent platform: model router, tool protocol (MCP), project memory (CLAUDE.md), subagents, hooks, skills. If you're already paying for Claude, Claude Code is the cheapest way to ship a real agent in-house.

The five extension points

Point	What it does	File location
CLAUDE.md	Persistent project/user memory (context on every turn).	~/.claude/CLAUDE.md, ./CLAUDE.md
MCP config	Connect tool servers.	~/.claude/claude_desktop_config.json
Subagents	Task-scoped agents with isolated context.	.claude/agents/.md, ~/.claude/agents/.md
Hooks	Shell commands on events (SubagentStart, PreToolUse, Stop).	.claude/settings.json hooks block
Skills	Folders of instructions + resources Claude pulls in as needed.	.claude/skills/*/

Custom subagent example

---
name: accessibility-reviewer
description: Reviews web UI for WCAG 2.2 compliance.
tools: Read, Grep, Glob, WebFetch
model: claude-sonnet-4-6
---

You are a strict WCAG 2.2 accessibility reviewer.

For any files or URLs given, check:
1. Color contrast ratios (4.5:1 body, 3:1 large).
2. Semantic HTML (headings, landmarks, labels).
3. Keyboard navigation (focus order, skip links, traps).
4. Alt text on images (descriptive, not filename).
5. Form labels and error states.

Report: issues as markdown table (severity / location / fix).
Never modify files. Read-only review.A subagent definition. Save as .claude/agents/accessibility-reviewer.md. Invoke from main Claude Code with /agents or @accessibility-reviewer.

Hooks for safety

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [{
          "type": "command",
          "command": "./.claude/hooks/block-rm-rf.sh"
        }]
      }
    ],
    "Stop": [
      {
        "hooks": [{
          "type": "command",
          "command": "osascript -e 'display notification \"Claude done\" with title \"CC\"'"
        }]
      }
    ]
  }
}Hooks: block dangerous bash commands, notify on task completion. Shell scripts the harness runs.

MCP in Claude Code

MCP servers declared in claude_desktop_config.json (global) or .mcp.json (project) become available as tools inside every Claude Code session and to every subagent (unless you restrict via the tools frontmatter field). This is how you give an agent Supabase, GitHub, or your private internal API access.

Skills — the agent-skills primitive

Skills are folders under .claude/skills/ or ~/.claude/skills/. Each skill has a SKILL.md with a description; Claude pulls it in when the description matches the user's task. Perfect for domain knowledge ('how our billing works', 'our code style guide') that shouldn't bloat CLAUDE.md.

# .claude/skills/deploy-runbook/SKILL.md

---
description: Use this skill when the user asks to deploy, release, or
push to production. Knows our specific deploy flow and gotchas.
---

# Deploy Runbook

Our deploy is a three-step GitHub Actions workflow:

1. Tag `vX.Y.Z` on main — triggers CI.
2. CI runs tests + builds artifacts.
3. Approval gate in GitHub Environments: prod requires a second reviewer.

Never run `kubectl apply` directly. Never edit prod secrets from CLI.
Rollback: `./scripts/rollback.sh <prev-tag>`.A skill definition. Claude uses it automatically when the description matches.

Common patterns

Project-specific subagents in .claude/agents/ (code reviewer, test writer, doc updater).
Org-wide subagents in ~/.claude/agents/ (security reviewer, deploy assistant).
Private MCP servers for internal APIs — auth via env vars in claude_desktop_config.json.
Hooks that post to Slack or a message queue on Stop (chain agents across sessions).
Skills for domain knowledge (style guides, runbooks) that loads lazily.

Claude Code is underrated as an agent platform. If your team is already on Claude, skip the framework selection rabbit hole for internal automation — use the CLI.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-agentic-claude-code-platform-creators

What is the core idea behind "Claude Code CLI as an Agent Platform"?
1. Claude Code isn't just a coding assistant — it's a general agent runtime with MCP, subagents, hooks, and skills. Treat it that way and you get a free, powerful platform.
2. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
Which term best describes a foundational idea in "Claude Code CLI as an Agent Platform"?
1. subagent
2. Claude Code
3. CLAUDE.md
4. hooks
A learner studying Claude Code CLI as an Agent Platform would need to understand which concept?
1. Claude Code
2. CLAUDE.md
3. subagent
4. hooks
Which of these is directly relevant to Claude Code CLI as an Agent Platform?
1. Claude Code
2. subagent
3. hooks
4. CLAUDE.md
Which of the following is a key point about Claude Code CLI as an Agent Platform?
1. Project-specific subagents in .claude/agents/ (code reviewer, test writer, doc updater).
2. Org-wide subagents in ~/.claude/agents/ (security reviewer, deploy assistant).
3. Private MCP servers for internal APIs — auth via env vars in claude_desktop_config.json.
4. Hooks that post to Slack or a message queue on Stop (chain agents across sessions).
Which of these does NOT belong in a discussion of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Project-specific subagents in .claude/agents/ (code reviewer, test writer, doc updater).
3. Private MCP servers for internal APIs — auth via env vars in claude_desktop_config.json.
4. Org-wide subagents in ~/.claude/agents/ (security reviewer, deploy assistant).
What is the key insight about "Subagents run up to 7 in parallel" in the context of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Spawning subagents via the Task tool lets Claude run independent threads concurrently — huge for codebase-wide searches …
4. Agent: 'Heads up: this took longer than expected.'
What is the key insight about "Hooks run your shell" in the context of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Agent: 'Heads up: this took longer than expected.'
4. Hooks execute real shell commands with your permissions. A malicious CLAUDE.md could add a hook that exfiltrates data.
What is the key warning about "Scope your agents tightly" in the context of Claude Code CLI as an Agent Platform?
1. Always define: goal, tools, permissions, and stop condition before executing.
2. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
Which statement accurately describes an aspect of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Claude Code ships as a developer CLI, but underneath it's a general agent platform: model router, tool protocol (MCP), project memory (CLAUD…
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
What does working with Claude Code CLI as an Agent Platform typically involve?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. MCP servers declared in claude_desktop_config.json (global) or .mcp.json (project) become available as tools inside every Claude Code sessio…
4. Agent: 'Heads up: this took longer than expected.'
Which of the following is true about Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Agent: 'Heads up: this took longer than expected.'
4. Skills are folders under .claude/skills/ or ~/.claude/skills/. Each skill has a SKILL.
Which best describes the scope of "Claude Code CLI as an Agent Platform"?
1. It focuses on Claude Code isn't just a coding assistant — it's a general agent runtime with MCP, subagents, hooks,
2. It is unrelated to agentic workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. The five extension points
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
Which section heading best belongs in a lesson about Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Custom subagent example
4. Agent: 'Heads up: this took longer than expected.'

← Back to interactive lesson

Tendril · Creators · Agentic AI

Claude Code CLI as an Agent Platform

Claude Code isn't just a coding assistant — it's a general agent runtime with MCP, subagents, hooks, and skills. Treat it that way and you get a free, powerful platform.

48 min · Reviewed 2026

More than a coding tool

The five extension points

Point	What it does	File location
CLAUDE.md	Persistent project/user memory (context on every turn).	~/.claude/CLAUDE.md, ./CLAUDE.md
MCP config	Connect tool servers.	~/.claude/claude_desktop_config.json
Subagents	Task-scoped agents with isolated context.	.claude/agents/.md, ~/.claude/agents/.md
Hooks	Shell commands on events (SubagentStart, PreToolUse, Stop).	.claude/settings.json hooks block
Skills	Folders of instructions + resources Claude pulls in as needed.	.claude/skills/*/

Custom subagent example

---
name: accessibility-reviewer
description: Reviews web UI for WCAG 2.2 compliance.
tools: Read, Grep, Glob, WebFetch
model: claude-sonnet-4-6
---

You are a strict WCAG 2.2 accessibility reviewer.

For any files or URLs given, check:
1. Color contrast ratios (4.5:1 body, 3:1 large).
2. Semantic HTML (headings, landmarks, labels).
3. Keyboard navigation (focus order, skip links, traps).
4. Alt text on images (descriptive, not filename).
5. Form labels and error states.

Report: issues as markdown table (severity / location / fix).
Never modify files. Read-only review.A subagent definition. Save as .claude/agents/accessibility-reviewer.md. Invoke from main Claude Code with /agents or @accessibility-reviewer.

Hooks for safety

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [{
          "type": "command",
          "command": "./.claude/hooks/block-rm-rf.sh"
        }]
      }
    ],
    "Stop": [
      {
        "hooks": [{
          "type": "command",
          "command": "osascript -e 'display notification \"Claude done\" with title \"CC\"'"
        }]
      }
    ]
  }
}Hooks: block dangerous bash commands, notify on task completion. Shell scripts the harness runs.

MCP in Claude Code

Skills — the agent-skills primitive

# .claude/skills/deploy-runbook/SKILL.md

---
description: Use this skill when the user asks to deploy, release, or
push to production. Knows our specific deploy flow and gotchas.
---

# Deploy Runbook

Our deploy is a three-step GitHub Actions workflow:

1. Tag `vX.Y.Z` on main — triggers CI.
2. CI runs tests + builds artifacts.
3. Approval gate in GitHub Environments: prod requires a second reviewer.

Never run `kubectl apply` directly. Never edit prod secrets from CLI.
Rollback: `./scripts/rollback.sh <prev-tag>`.A skill definition. Claude uses it automatically when the description matches.

Common patterns

Project-specific subagents in .claude/agents/ (code reviewer, test writer, doc updater).
Org-wide subagents in ~/.claude/agents/ (security reviewer, deploy assistant).
Private MCP servers for internal APIs — auth via env vars in claude_desktop_config.json.
Hooks that post to Slack or a message queue on Stop (chain agents across sessions).
Skills for domain knowledge (style guides, runbooks) that loads lazily.

Claude Code is underrated as an agent platform. If your team is already on Claude, skip the framework selection rabbit hole for internal automation — use the CLI.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-agentic-claude-code-platform-creators

What is the core idea behind "Claude Code CLI as an Agent Platform"?
1. Claude Code isn't just a coding assistant — it's a general agent runtime with MCP, subagents, hooks, and skills. Treat it that way and you get a free, powerful platform.
2. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
Which term best describes a foundational idea in "Claude Code CLI as an Agent Platform"?
1. subagent
2. Claude Code
3. CLAUDE.md
4. hooks
A learner studying Claude Code CLI as an Agent Platform would need to understand which concept?
1. Claude Code
2. CLAUDE.md
3. subagent
4. hooks
Which of these is directly relevant to Claude Code CLI as an Agent Platform?
1. Claude Code
2. subagent
3. hooks
4. CLAUDE.md
Which of the following is a key point about Claude Code CLI as an Agent Platform?
1. Project-specific subagents in .claude/agents/ (code reviewer, test writer, doc updater).
2. Org-wide subagents in ~/.claude/agents/ (security reviewer, deploy assistant).
3. Private MCP servers for internal APIs — auth via env vars in claude_desktop_config.json.
4. Hooks that post to Slack or a message queue on Stop (chain agents across sessions).
Which of these does NOT belong in a discussion of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Project-specific subagents in .claude/agents/ (code reviewer, test writer, doc updater).
3. Private MCP servers for internal APIs — auth via env vars in claude_desktop_config.json.
4. Org-wide subagents in ~/.claude/agents/ (security reviewer, deploy assistant).
What is the key insight about "Subagents run up to 7 in parallel" in the context of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Spawning subagents via the Task tool lets Claude run independent threads concurrently — huge for codebase-wide searches …
4. Agent: 'Heads up: this took longer than expected.'
What is the key insight about "Hooks run your shell" in the context of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Agent: 'Heads up: this took longer than expected.'
4. Hooks execute real shell commands with your permissions. A malicious CLAUDE.md could add a hook that exfiltrates data.
What is the key warning about "Scope your agents tightly" in the context of Claude Code CLI as an Agent Platform?
1. Always define: goal, tools, permissions, and stop condition before executing.
2. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
Which statement accurately describes an aspect of Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Claude Code ships as a developer CLI, but underneath it's a general agent platform: model router, tool protocol (MCP), project memory (CLAUD…
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
What does working with Claude Code CLI as an Agent Platform typically involve?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. MCP servers declared in claude_desktop_config.json (global) or .mcp.json (project) become available as tools inside every Claude Code sessio…
4. Agent: 'Heads up: this took longer than expected.'
Which of the following is true about Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Agent: 'Heads up: this took longer than expected.'
4. Skills are folders under .claude/skills/ or ~/.claude/skills/. Each skill has a SKILL.
Which best describes the scope of "Claude Code CLI as an Agent Platform"?
1. It focuses on Claude Code isn't just a coding assistant — it's a general agent runtime with MCP, subagents, hooks,
2. It is unrelated to agentic workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. The five extension points
3. Replace a real security review.
4. Agent: 'Heads up: this took longer than expected.'
Which section heading best belongs in a lesson about Claude Code CLI as an Agent Platform?
1. Tell it your idea — agent asks 'What's the smallest version that's still useful?…
2. Replace a real security review.
3. Custom subagent example
4. Agent: 'Heads up: this took longer than expected.'

← Back to interactive lesson