LZLZL/AI toolchain/Always-on · Tools
FREEMedium · Practice C · Always-ontool surface

MCP and the tool surface:
what not to install

2026-08-21 · The tool surface is the risk surface

MCP is an open protocol letting an agent reach outside tools and data. The more you connect, the more it can do — and the more dangerous it is. Here is how to grade tools by risk, and one principle that beats everything: the minimum tool set.

In this pagewhat MCP is, risk tiers, the cost of third-party servers. Not in this pagewhich servers I run.

1What MCP is

In one line: an open protocol for a model to talk to outside tools.

Before it, every tool needed adapting to every agent. With it, each side implements once — a server (the capability: a database, an API, a filesystem) and a client (the agent). Any combination works.

In practice, what you feel is: install an MCP server, the agent gains a set of abilities. Which is exactly where the risk comes from — it is too easy.

2Three tiers, graded by "can it send things out"

There are many ways to grade tools; this one is the most useful, because it maps directly onto how irreversible the damage is:

TierCapabilityWorst caseDefault stance
🟢 Read-onlyRead files, query databases, searchIt read something it should not (still on your machine)Fine to enable
🟡 Local writeEdit files, run commands, change configBroken local state. Version control undoes itDepends; configure permissions
🔴 OutboundSend messages, email, call external APIs, move money, publishIrreversible, and possibly on someone elseOff by default
Why red gets its own tier

Mistakes in the first two stay on your machine. You can roll back and retry. Once the third executes, the thing has left — message sent, money moved, post public.

And it is the tier most likely to get installed casually: a notification hook, an email connector, an auto-publisher. Each looks convenient on its own.

Recommended stance: keep a human confirmation on red capabilities, even at the cost of full automation. That is page 10's third test — can you absorb it being wrong — made concrete.

⚠ 🟢 plus 🔴 is worse than 🔴 alone

Read-only tools pull sensitive content into context; outbound tools send context out. Each looks acceptable alone; together they form a complete exfiltration path.

That is what makes the previous page's second threat layer genuinely dangerous: a fooled agent holding both categories can be talked into shipping your secrets in one sentence.

3⚠ A third-party server is someone else's code on your machine

Worth reading twice

An MCP server is a real program, running on your machine, usually with whatever permissions you gave it. Installing one is, in trust terms, identical to installing an unfamiliar CLI tool and running it.

And it carries one extra layer: its tool descriptions enter the model's context — text written by the server's author is read by the model as instruction-shaped content.

Three defences:
① prefer ones that are open source and traceable;
② check what permissions it asks for — a weather tool wanting your home directory is wrong;
③ run anything uncertain in an isolated environment first (Docker or a remote sandbox — page 17).

4The minimum tool set: more is not just riskier, it is dumber

The part people miss: every enabled tool spends context on its name and description.

ConsequenceWhy
More expensiveTool definitions are fixed overhead resent every turn (page 9)
Less accurateMore options, more chance of the wrong pick. One-of-twenty goes wrong more often than one-of-five
Enable per project, not globally

Most tools can be toggled at will (Hermes uses hermes tools). The right posture is a small standing set, extended when needed, not "it does not hurt to leave everything on".

Self-check: can you list, from memory, what is currently enabled? If not, you have too many.

5A minimal starting set

YesNot yet
Read files / searchAnything that sends messages or email
Run commands (with deny rules set)Anything touching payments, transfers or orders
Version control operationsAnything that publishes content
Fetch web pages (read-only)Anything that can change your account settings

The right column is not "never" — it is "once you have a feel for how this behaves". Two weeks, then decide what to add.

MCP An open protocol connecting models to external tools and data, split into servers (capability) and clients (agent); supported by the mainstream agents including Hermes.
Toggling Hermes enables and disables tools with hermes tools. Official docs, checked 2026-08-21.
Context overhead Tool names and descriptions are fixed per-turn overhead; mechanism on page 9.
Nature of this page The three-tier split and the minimum-tool-set principle are this site's practice and judgement, not any official specification. Adjust to your own risk tolerance.
Not in this page Which tools or MCP servers I have enabled.

RelatedRead next

C · Always-on
Telegram: lock down who, then discuss what
D · Cloud
Move only the hands: the SSH sandbox
A · Local
Claude Code: install to first real edit
D · Cloud
Key isolation, monitoring, backups, switching off
An educational and engineering record — not a review or endorsement of any third-party product. Commands, config keys, prices and terms are per each vendor's official docs; this page states when it was checked and all of them can change without notice — verify before you copy anything. Self-hosting is your own responsibility: keys, accounts and data are on you.