8.3 Agentic customization: built-in vs custom vs Skills vs MCP
The four options
| Option | What it is | You maintain | Reuse across apps |
|---|---|---|---|
| Built-in / server-side tools | Anthropic-provided capabilities: web search, web fetch, code execution, bash, text editor, memory; in Claude Code also Read/Write/Edit/Grep/Glob | Nothing | Automatic |
| Custom tools | Your function definitions plus handlers in your process | Schema, handler, tests | Only by copying code |
| Skills | A packaged procedure — instructions, optionally scripts and restricted tools — loaded on demand | A markdown file (plus assets) | Via the repo or a plugin |
| MCP servers | A separate process exposing tools, resources, and prompts over a protocol | A service with its own lifecycle | Across every MCP client |
Choosing
| Requirement | Choose |
|---|---|
| Public web information | Built-in web search / web fetch — do not write your own scraper tool |
| Run code, compute, transform data | Built-in code execution |
| Read and edit files in a repo | Built-in Claude Code tools |
| Reach your database, internal API, or SaaS account | Custom tool or MCP server |
| One application needs it, nobody else | Custom tool — least infrastructure |
| Several applications need it, maintained independently | MCP server |
| Encode how your team performs a recurring task | Skill — it is procedure, not capability |
| A capability plus a procedure | Both: a tool for the capability, a skill that describes the workflow |
| Large tool inventory, most irrelevant per request | Tool search with defer_loading, or split by subagent role |
A tool is a capability — something the agent can do that it otherwise cannot. A skill is a procedure — instructions for doing something with capabilities it already has. "The team keeps doing this five-step review differently" is a skill. "Claude cannot see our inventory levels" is a tool or an MCP server. Answering "write a skill" to a capability gap, or "write a tool" to a consistency gap, is the mistake being tested.
What each option costs
| Option | Token cost | Latency | Operational cost | Risk |
|---|---|---|---|---|
| Built-in | Definition in the prefix; results can be large | Server-side execution | None | Output is untrusted external data |
| Custom tool | Definition in the prefix | Your handler's latency | Your code, your tests | You own correctness and authorization |
| Skill | Zero until invoked | Its own turns | A file to keep current | Stale procedure silently followed |
| MCP server | Definitions in the prefix, per connected server | Process or network hop | A service to run, version, and secure | Untrusted output; a shared blast radius |
Two consequences worth memorising: connecting many MCP servers is not free — every tool definition sits in the prefix of every request — and a skill costs nothing when unused, which is why procedure belongs in a skill rather than in CLAUDE.md.
Combining them
A realistic support agent:
| Layer | Choice |
|---|---|
| Look up customers and orders | MCP server in front of the internal APIs — reused by the support agent, the ops dashboard agent, and Claude Code |
| Issue a refund | Custom tool in the application process, so the approval gate and the caller's authorization stay in one place |
| Check public carrier status | Built-in web fetch |
| Summarise the shift's escalations | Skill — the procedure the team agreed on |
| Enforce the refund cap | Hook — not a tool at all |
The last row is the point: the customization question and the enforcement question are separate. Adding a tool never enforces a rule about that tool.
Quiz
- Question 1
Reviewers apply the team’s five-step code-review procedure inconsistently. Claude already has every tool the review needs, and the procedure should run identically whether a reviewer asks for it by name or simply asks Claude to review a pull request. What should be built?
- Question 2
An agent needs current public information from the web, must run Python to compute a projection, and must read internal sales figures from a private warehouse. Which combination is correct?
Links
- https://platform.claude.com/docs/en/agents-and-tools/tool-use/overview
- https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview
- https://platform.claude.com/docs/en/agents-and-tools/mcp
- https://code.claude.com/docs/en/skills
- https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills
- https://www.anthropic.com/engineering/writing-tools-for-agents