Mann vs Machine on Community Servers

Mann vs Machine is Team Fortress 2 played co-operatively against waves of robots, and it is the only mode in the game where six strangers have to agree on a plan before anything works.

Valve shipped it in 2012 with a handful of missions. The community has been running it ever since, on servers with custom missions, adjusted difficulty and conventions that nobody writes down.

How a mission works

Six players defend a point against waves of robots walking a fixed path carrying a bomb. If the bomb reaches the hatch, you lose the wave.

Between waves you spend money, dropped by robots — on upgrades. Damage, firing speed, resistances, healing, and class-specific abilities. Upgrades persist through the mission and reset at the end of it.

A mission is five to nine waves and runs thirty to sixty minutes. Losing a wave restarts that wave, not the mission, which is what makes the format tolerant of failure in a way most co-op is not.

Why it needs coordination

The mode is balanced around a team covering several jobs at once, and a team where everybody bought damage upgrades loses to wave four every time.

Somebody has to handle the bomb carrier. Robots that pick up the bomb gain defensive buffs as they advance. Ignoring this is the most common way a wave is lost.

Somebody has to handle giants. Concentrated damage, usually a Heavy or a Soldier with the right upgrades.

Somebody has to keep the team alive. A Medic is close to mandatory on harder missions.

Somebody has to handle the swarm. Pyros and Demos clearing the small robots so the rest is manageable.

None of this is assigned by the game. It is negotiated in the first thirty seconds, or it is not, and whether it happens is the whole difference between a good session and forty minutes of failing the same wave.

Upgrading well

The part new players get wrong, and it is not intuitive.

Buy resistances before damage on hard missions. Surviving is the constraint, not output.

Buy movement speed early on classes that need to reposition. Underrated and cheap.

Do not spread upgrades thin. Two maxed abilities beat six partial ones.

Money is shared in effect. Robots drop credits everyone can collect, and picking them all up matters: missed credits are a permanent reduction in the team's total for that mission.

Look at what the team lacks. If nobody has bought healing, buying more damage is not the answer.

Most servers have players who will tell you what to buy if you ask, and asking is normal rather than embarrassing.

Official against community

Valve's own MvM runs through matchmaking with a fixed mission set, and Mann Up mode has item rewards attached.

Community servers run custom missions, and this is where the mode is actually alive. Missions built by players, difficulties beyond what Valve shipped, custom robots, and rotations that go far past the official six maps.

The custom mission scene is the reason the mode persisted. Like custom campaigns in Left 4 Dead 2, the community produced more content than the developer did, and it kept producing it after the developer stopped.

The TF2 community servers guide covers the wider ecosystem, and the TF2 listings are where to find them.

Choosing a server

Difficulty tier. Community missions run from easier than official to substantially harder. Servers usually state it and the labels are not standardised, so a first visit is reconnaissance.

Whether it is beginner-friendly. Some servers are explicitly for learning; others assume you know the upgrade meta. The second kind is not hostile, it is just fast.

Population at your hour. Six players is a full team and the mode does not work with four. A server showing two is a server you cannot play on. Peak hours is the check.

Expect a download. Custom missions and robots are content, and a server with a deep rotation sends a lot on first join. Why modded servers take so long to join.

Playing with strangers

The mode is unusually tolerant of new players, for a specific reason: a wave lost is a wave repeated, so a mistake costs minutes rather than the session.

Say you are new. Nearly always produces advice rather than annoyance.

Ask what class the team needs before picking.

Do not switch class mid-mission without saying so, upgrades are per-class and switching throws yours away.

Stay for the whole mission. Leaving at wave six means the remaining five play a wave balanced for six. This is the one thing that irritates regulars.

Running one

Population floor is exactly six, which is the format's main operational problem. Below it the mode does not function, and it does not degrade gracefully; this is the same structural issue jailbreak and zombie escape have.

Mission curation is the product. A rotation weighted towards approachable missions keeps new players; one full of the hardest community content is a server for the same eight people.

Serve content properly. Custom missions, robots and maps add up.

Moderation is light. Co-op, nothing persistent, and the main friction is people leaving mid-mission.

The plugin stack is mature and stable. TF2 updates rarely now, which makes this a low-maintenance server compared with anything on an actively developed game. Updating without breaking things still applies, less often.

The classes, and what each is for here

Different from the rest of Team Fortress 2, because the opponents do not think and the constraints are entirely different.

Heavy is the backbone of most compositions. Sustained damage into a single point, enormous health with resistances bought, and the class most able to hold a choke against giants.

Medic is close to mandatory above the easier difficulties. The upgraded shield and the healing throughput are what let a team survive waves designed to overwhelm one. A team without a Medic is a team betting on killing everything before it arrives.

Demoman clears crowds better than anything else and does it from safety. On many missions this is the highest-value slot and it is frequently empty because people want to shoot giants.

Soldier is flexible: burst damage for giants, splash for crowds, and the class most able to reposition after upgrades.

Scout collects money. This is not a joke role. Credits missed are credits gone for the mission, and a Scout with movement upgrades collects everything while the rest hold the line. Also the fastest way to slow a bomb carrier.

Engineer contributes a teleporter that saves the walk back and a sentry that holds a flank. Underrated on long missions.

Pyro handles the swarm and can push a bomb carrier away from the hatch, which occasionally saves a wave that was lost.

Sniper and Spy are situational and are a liability on a team of six that needs coverage. Both are viable on specific missions with players who know them.

Where missions come from

Worth knowing, because the mission is the product and the server is the venue.

Missions are made by players using the game's own map and mission tools, and the good ones are curated collections rather than one-off files. A server with a deep rotation is running somebody's judgement about what is worth playing, and that curation is the difference between servers more than the hardware is.

Difficulty labels are not standardised. What one server calls advanced another calls intermediate, and the community's own naming conventions have shifted over the years. Treat the label as a hint and the first wave as the measurement.

Custom robots are common in community missions: new weapons, new behaviours, giants that behave differently from Valve's. This is where the format stayed interesting and it is also why a first visit to an unfamiliar server can be surprising.

Some missions are built to be nearly impossible, deliberately, as a challenge for groups who have exhausted everything else. Joining one of those by accident is a specific and memorable experience.

Running one, practically

The operational details behind the shorter version above.

Content delivery is the first thing to get right. Custom missions, robots and maps have to reach the client, and a server relying on the slow default path gives every new visitor a long wait before they have seen anything. A fast download host fixes it and is the highest-return change most MvM servers can make. Why modded servers take so long to join covers what players are waiting through.

Rotation weighting is your retention strategy. Front the approachable missions, keep the hard ones available, and resist the pull towards curating for your own regulars, who are the eight people who will stay regardless.

A voting system for the next mission works better here than in most formats, because the group that just played together has a shared opinion about what to do next.

Population is the whole problem. Six is the floor and the mode does not degrade below it, so the practical answer is concentrating on fewer server instances at known hours rather than running several that are each half full. When servers are actually busy covers finding the hour worth defending.

Worth trying

MvM is the most cooperative thing in Team Fortress 2 and one of the few modes anywhere where six strangers have to talk to each other to succeed.

Find a beginner-friendly server at its peak, say you are new, ask what to buy, and stay to the end. That is a complete first session, and the format is forgiving enough that being bad at it costs nobody anything.