Terminal Trenches Guide: Filters, Alerts and Safety Signals (2026)

Trenches is the surface Terminal is actually built around. Not the charts, not the portfolio page. The production bundle references it 408 times, more than any other feature, and once you have used it for an hour it is obvious why: this is where the product's opinion about what trading looks like lives.

It is also the surface where the honest framing matters most. Around 0.26 percent of pump.fun tokens graduated as of June 2026, and CoinGecko measured a 69 percent launch-day death rate in the same month. No feed, no filter set, and no safety panel changes that arithmetic. What good filtering does is cut down how many obvious traps you personally walk into, which is a smaller claim than most tools make and the only one supported by the numbers.

The three columns

Trenches splits into New, Almost Bonded and Recently Bonded, and each column is a different game.

New is raw launch flow. Almost nothing here survives. This column is where the safety signals earn their keep, because a token twenty seconds old has no chart, no volume history, and nothing to judge except who deployed it and how the supply was distributed at birth.

Almost Bonded is tokens approaching the end of the bonding curve. There is real information here: enough buy pressure has accumulated to push a token most of the way up the curve, and holder counts and volume have started to mean something. It is also the most crowded column, because everyone else can see the same progress bar.

Recently Bonded is post-migration. The token now trades on a real pool, and the crowd that only buys graduated tokens has arrived. Note that bundle analysis stops updating at migration, so the risk read you would have used in the first two columns is partly gone here.

One thing to know if you followed an old documentation link: there is no /new-pairs route in the current app despite the docs pointing at one. Trenches replaced it.

Which launchpads feed it

The archived documentation lists two sources: pump.fun and Four.meme. The live app supports considerably more. Reading the venue list out of the production bundle, Trenches also covers Trench, Flap, Virtuals, Circus and Hood.fun.

Hood.fun is the interesting one, because it is the launchpad on Robinhood Chain, which is fully wired into Terminal (explorer, quote tokens, chain color, the lot) and appears in exactly none of the marketing or documentation. Robinhood has publicly denied involvement in the pump.fun integration. Make of that what you will, but the chain selector ships five networks while the marketing claims four.

The filter set

This is where the time investment pays off. Filters are saved as named presets, so you build a thesis once and reuse it. The full set of numeric ranges available:

Plus three toggles that are simple and do a lot of work: Dex paid, has socials, and dev still holding.

Dex paid means someone spent money on a DEX Screener listing upgrade, which is weak evidence of intent but real evidence that somebody was willing to spend. Has socials filters out the completely anonymous launches. Dev still holding is the one I would think hardest about, because it cuts both ways: a deployer holding nothing has already exited, and a deployer holding a lot can exit whenever they want.

Card metrics and display options

Around 19 metrics and display options can be toggled on and off per token card. Because Terminal's interface is extremely dense already (24 pixel row heights, tabular mono numerals, 12 to 13 pixel body text), turning everything on produces a wall you cannot scan.

The version I would build: keep market cap, age, holder count, top-10 concentration, dev holding and the volume figure. Everything else goes off. The point of a feed is that your eye catches an anomaly in under a second, and that only happens when there are few enough numbers per card for a pattern to exist.

Sound alerts

Attach a sound to a filter and Trenches makes a noise when something matches. The available sounds are a real enum in the code, and the list is exactly as unserious as it looks: CORK, RING, BELL, NICE, YEAH BOY, KACHING, UGH and CUSTOM.

CUSTOM is the one that matters. Distinct sounds mapped to distinct filters is the only way this scales, because a single alert tone across three saved filters tells you nothing except that something happened. Map your tightest filter to something you would never confuse with anything else, and leave your loose exploratory filters silent.

Small warning from experience with any alert system: the noise trains you. If your alert fires forty times an hour, you stop hearing it inside a day and the feature is worse than useless because you now believe you have coverage you do not have.

The dev blacklist

Up to 200 addresses. Blacklisted deployers disappear from your feed.

Two hundred is a strange number in context. Pump.fun was minting roughly 30,000 tokens a day in January 2026, and serial deployers rotate wallets constantly, so a 200-slot list is nowhere near a comprehensive defense. What it is good for is memory. The deployer who rugged you personally, the one whose naming convention you have now seen four times, the wallet you traced from a funding check. Those go on the list and stay off your screen forever, which is worth more than it sounds when you are scanning fast and tired.

The safety signals, and what each one tells you

This is the part of Trenches that justifies using it over just watching pump.fun's own site. Each signal answers a narrow question, and knowing which question is the difference between a filter and a superstition.

Bundle analysis

Shows how much of the supply was acquired in the same transaction bundle as the launch. In practice that means the deployer and whoever they coordinated with bought before anyone outside could place a bid.

A high bundled percentage means a large block of supply sits with people who got it at effectively zero cost and can exit at any price above that. It is the strongest single pre-migration risk signal in the product.

It has limits. Bundle analysis is Solana only, and it stops at Raydium migration, so in the Recently Bonded column you are working without it.

Dev holding percentage

How much the deployer kept. Both extremes are informative. Near zero often means they have already sold, and the question becomes who bought. A large retained share means the largest possible seller is still loaded and has not moved yet.

There is no good number here, which is why this signal only works alongside the others. Read it together with dev funding and bundle percentage rather than on its own.

Insider holding percentage

Supply concentrated in wallets Terminal has associated with the launch, which is a wider net than the deployer address alone. It catches the pattern where a deployer holds almost nothing while five wallets funded from the same source hold 40 percent between them. That structure is designed to make the dev holding number look clean.

Dev funding source and timing

Where the deployer's wallet got its SOL, and when. This is the most underrated signal in the panel.

A deployer wallet funded twenty minutes before launch, directly from a centralized exchange withdrawal, with no other history, is a wallet created for this launch. A wallet funded from another deployer wallet you have seen before is a serial operator, which the blacklist exists for. Funding history is hard for a deployer to fake because it is a chain of real transactions with real timestamps.

Fresh wallet buys

Counts buyers whose wallets are new and have no meaningful history. A launch where most of the early volume comes from wallets minted the same day is not attracting a market, it is displaying one.

Solana only, like most of this panel.

Bot transaction detection

Terminal flags transactions it recognizes as coming from trading bots, naming Maestro and Photon among them. Bot flow is not automatically bad, plenty of ordinary traders route through bots. What it tells you is who you are trading against and how fast they are. A launch that is overwhelmingly bot flow is a market where your click speed is the product being sold.

Avatar reused

Flags a token whose image has been used by a previous token. Cheap to check, and it catches the laziest category of repeat scammer, the one recycling the same asset pack across launches. Not conclusive on its own, since plenty of legitimate tokens use stock imagery, but combined with a fresh deployer wallet it is close to a decision.

Liquidity lock icon

Indicates whether pool liquidity is locked. Unlocked liquidity on a migrated token means it can be pulled. This is a straightforward binary and it matters most in the Recently Bonded column, where bundle analysis has gone dark.

Freeze authority warning

Flags tokens where the freeze authority has not been revoked, meaning the token's authority can freeze your account and prevent you from selling. Terminal has a specific error string for this outcome: "The developer of {token} froze your account, preventing you from trading this token."

That string exists in the shipped app because it happens. This is the one signal on the list I would treat as an automatic no rather than a data point, since the downside is not a bad price, it is a position you cannot exit at any price.

Token tax display

Shows buy and sell taxes on the contract. Common on EVM tokens, and the number to watch is the asymmetry: a 5 percent buy tax with a 40 percent sell tax is a honeypot with extra steps. Terminal also runs honeypot.is checks on EVM chains and offers an EVA Audit AI contract checker, though that last one is Ethereum and Base only.

How to set it up

That fourth step is not a throwaway. It is the single most costly setting on the platform for anyone trading small size, and the full write-up is here: why small trades lose money on Terminal.

What Trenches will not do

It will not find you a winner. With a graduation rate near a quarter of a percent and most tokens dead on day one, the base rate is doing far more work than any filter you can build. Discovery tools sell the feeling of an edge, and the honest version of the pitch is narrower: Trenches makes it fast to reject things, and rejection at speed is worth a lot when 30,000 tokens launch a day.

It also does not replace an exit plan. Getting in early on something that runs is only half of it, and the reason people give back gains is that they had no structure for selling. Terminal is unusually well equipped there, with position-sized take-profits and up to five exit conditions attached at entry, and that is covered in full in our order types and exit strategies guide.

Also worth pricing in: Terminal charges a flat 1 percent trading fee with no volume tiers, stacked on top of pump.fun's own 1.25 percent bonding-curve fee, which is a lot of friction when you are taking many small shots at low-cap tokens. The fee breakdown shows how that compounds. And the cashback rate on those fees is set permanently at signup, which the referral page explains.

The bottom line

Trenches is the best reason to use Terminal, and the safety panel is better than what most competing feeds show you, particularly the funding-source and bundle checks. If you are comparing against the platform taking the most share right now, our Terminal vs Axiom comparison covers execution speed, which is where Terminal draws its most consistent criticism.

Use the feed to throw things out. Keep the filter set small enough that you actually read the results. And do not confuse a token that passed your filters with a token that is going to work, because those two things are separated by a 0.26 percent graduation rate.

You can open the feed at terminal.pump.fun, and if you want the background on who runs the platform and how it got here, start with what Terminal is.