# MCP Server: The Tool List Is the Capability List

What an MCP server exposes — tools, resources, prompts — and how to read its tool list line by line to see what an AI can and cannot do through it.

Glossary · Published 2026-08-10, updated 2026-10-22.

> An MCP server is a program that exposes a defined set of tools — named actions with typed inputs — that an AI assistant can call over the Model Context Protocol. It sits between the AI client and a real system such as the Google Ads API, executes only the actions it defines, and returns structured results. Its tool list is a complete, readable statement of what the AI can and cannot do.

An MCP server is a program that exposes a menu of actions — tools — that an AI assistant is allowed to order from. Nothing more mystical than that. Each tool has a name, a description the model reads, and a schema for its inputs. The server sits between the AI client and a real system, executes the calls it receives, and returns results as structured data. It is the entire boundary between the AI and your stuff.

That framing has a useful consequence: the tool list is the capability list. Read it and you know precisely what the AI can do through the server — and, just as important, what it cannot.

## How an MCP server works

The [Model Context Protocol](https://adcopilot.cloud/glossary/model-context-protocol) defines three things a server can expose:

- **Tools** — actions the model may request: run a query, create a campaign, add a negative keyword. This is where nearly all the value and all the risk live.
- **Resources** — readable data the client can load as context: files, reports, documents.
- **Prompts** — pre-written instruction templates the server offers the user.

A call is a round trip. The client sends a tool name plus arguments; the server validates them, executes against the underlying system, and returns the result; the model reads it and continues. The server never initiates anything — no schedule, no background jobs. If nobody asks, nothing runs.

### Local or remote — the same server, a different address

A server can run on your machine, launched by the client with credentials in local config files, or on hosted infrastructure, reached as a URL. The protocol is identical; the operational burden is not. Who patches it, who stores the tokens, who can revoke access — those answers differ enough that [remote versus local](https://adcopilot.cloud/glossary/remote-mcp-server) deserves its own entry.

## What a Google Ads MCP server's tool list looks like

Concretely: as of v2.46.25, AdCopilot's underlying server implements 82 Google Ads tools, 63 in its fork of Google's MCP server and 19 Google Ads tools of its own (read-only analyses, and one write that rates a Local Services lead), which is about what a full Google Ads MCP server runs to. Reads — the list of accounts your sign-in can reach, GAQL queries for spend, search terms, change history, and audits that work out their findings in code. Writes — create search and Performance Max campaigns, ad groups, responsive search ads, assets, keywords and negative keywords, conversion actions, budgets, bidding strategies, geo targeting, ad schedules, pause and enable.

What matters just as much is what a careful hosted deployment leaves out. AdCopilot's [hosted connector](https://adcopilot.cloud/google-ads-mcp-server) exposes 76 of those 82 Google Ads tools (as of v2.46.25). The five `remove_*` tools that delete a campaign, an ad, a keyword or an asset are never exposed, and any change carrying status REMOVED is refused on the server before it reaches Google — in any letter case. Deletion is not gated behind a prompt; the capability is absent. That distinction only exists because a server's tool list is an enforceable boundary, not a suggestion.

## Reading a tool list, line by line

Take four lines from the list AdCopilot's hosted connector serves, in plain words, and see what each one settles:

| A line in the list | Kind | What it settles |
|---|---|---|
| Lists the accounts this sign-in can reach | Read | The server starts from the access the signed-in person already has, not from a master key. |
| Runs a query against account data | Read | Any report is possible, and no report changes anything. |
| Adds negative keywords to a campaign | Write | A change is possible, but this one adds a block rather than removing anything, and it can be taken back out. |
| Pauses or enables a campaign | Write | A status flip: the campaign and its history stay where they are. |

Then read what is not there: no line deletes a campaign, an ad group, an ad, a keyword or an asset. The only removals on the list, four of them, take negative keywords back out (from a campaign, from the account, or from a shared list) or take a location off a campaign, which is how what the third line adds can be removed again. No prompt gates the missing lines and no setting switches them on; ask the server for one and the answer is that no such tool exists. A list that is mostly reads is a different server from one that is mostly writes, and a list that does carry delete lines is a different server again. Reading it takes a minute and needs no trust in what the vendor says about itself.

## How to judge any server in five minutes

The tool list turns vendor evaluation from marketing-reading into an inspection. Five minutes with any MCP server's list answers five questions:

1. **What can it read?** Enough to be useful — search terms, spend, change history?
2. **What can it write?** Campaigns and keywords only, or budgets and statuses too?
3. **Can it delete?** If remove tools are present, ask why. If they are absent, that is architecture doing safety's job.
4. **Where do credentials live?** In local files you secure, or in [hosted custody you can revoke](https://adcopilot.cloud/security)? That choice has [its own entry](https://adcopilot.cloud/glossary/remote-mcp-server).
5. **What gets logged?** Every call, or only writes?

An AI managing your account is exactly as capable, and exactly as bounded, as the server you connect it to. Choose the server the way you would set a new hire's permissions on day one — because that is literally what it is. Start with [what MCP itself is](https://adcopilot.cloud/glossary/model-context-protocol), then read a [live Google Ads server](https://adcopilot.cloud/google-ads-mcp-server) with this checklist in hand.

## Questions

### Is an MCP server the same as an API?

No — it wraps one. The underlying API (Google Ads, a database, a file system) does the real work. The MCP server translates that API into tools an AI can discover and call: each tool has a name, a plain-language description and a typed input schema. The API speaks to programs written in advance; the server lets a model decide at runtime which call fits the conversation.

### Can an MCP server act without being asked?

No. A server only responds to requests from a connected client — it has no schedule, no background jobs, no will of its own. If nobody asks, nothing runs. Recurring work in MCP setups comes from the client side, a user or an agent loop initiating calls, which is also where approval prompts for write actions live.

---

This is the markdown copy of https://adcopilot.cloud/glossary/mcp-server. The index of the site's pages is https://adcopilot.cloud/llms.txt.
