00 read this with your agent
Read this with your agent first.
Copy one of these, paste it into whatever AI you use, and it will teach you from this page and only this page, at your level. An agent in a browser that supports WebMCP can use the page’s tools directly.
01 the seed in the fruit
Six layers, and the one you drink is the innermost.
Everything that happens after the harvest is a decision about when to take which layer off.
A ripe coffee cherry is the size of a large blueberry and the colour of a cranberry. Cut one across and you find six layers. Get this picture into your head and the rest of the post follows from it.
From the outside in: the skin, a thin shiny case; the pulp, fruit in the ordinary sense, sweet and edible; the mucilage, a thin sticky gel with sugar dissolved through it; the parchment, a papery hull that dries hard; the silverskin, a fine coat on the seed; and the seed, which is what we call a coffee bean. Most cherries hold two of them, flat faces pressed together.
By weight, the skin and pulp come to about 43% of the fresh fruit, the mucilage and its sugars about 12%, and the parchment about 6%1. The seeds are what is left: well under half. Most of a coffee cherry is not coffee, and processing is the business of getting rid of the rest of it without ruining what is left.
The mucilage matters far more than its share of the weight suggests, because that is where the sugar is. Spread across every bean on a drying bed, in the tropical heat, it is a banquet, and something is always going to come and eat it.
For the enthusiast: the botanists' names, and what the mucilage is made of
The skin is the exocarp and the pulp the mesocarp. The mucilage is mostly pectin, the same family of compounds that sets jam. The parchment is the endocarp. The silverskin is the seed coat, and it clings so tightly that scraps of it still flake off in the roaster. The seed is the endosperm: each one a half dome with a fold running down the flat side.
Analyses put coffee mucilage at around 84% water, 9% protein and 4% sugar by weight, with pectic substances making up most of the remainder2. Four percent sounds like nothing. It is the accessible sugar, and it is what everything in section 03 runs on.
| Layer | Also called | What it is |
|---|---|---|
| skin | exocarp | The outer case. Red, purple or yellow when ripe, depending on the variety. |
| pulp | mesocarp | The flesh of the fruit. Sweet and edible; most of it is thrown away or composted. |
| mucilage | pectin layer | A sticky gel bonded to the parchment. Pectin, water, sugar. The fuel for everything in section 03. |
| parchment | endocarp | A papery hull that dries to a brittle shell. Hulled off at the mill, not on the farm. |
| silverskin | seed coat, testa | A fine skin on the seed itself. What becomes chaff in the roaster. |
| the seed | endosperm | The coffee. Alive, respiring, and capable of germinating right up until it is dried. |
02 the fork in the road
Four answers to one question: when does the fruit come off.
Washed takes the fruit off at once, honey leaves the sticky layer on, and natural leaves all of it on until the end.
Washed takes the skin and pulp off within hours and breaks the mucilage down in a tank: a short, wet fermentation, ended on purpose. Honey, or pulped natural, dries the bean with the mucilage still on. Natural, the dry process, puts the whole cherry on a bed and hulls everything away weeks later, when it is brittle. Wet hulling, the Indonesian variant, removes the parchment while the bean is still wet11.
Each of the others has its own post. From here on, this one is about the dry route only.
- coffee/washed-processthe wet route
- coffee/honey-processthe middle route, and the experimental practices
- coffee/fermentationthe microbiology, across all of them
03 what happens on the bed
The fruit ferments where it lies, sealed inside its own skin.
A tank is a bath. A drying bed is a shrinking, sealed vessel, one of them per cherry.
In a washed coffee the beans sit loose in water, and somebody drains the tank when the mucilage lets go. In a natural nothing is exposed and nothing is removed: each cherry is its own small vessel, and whatever is going to eat that sugar has to do it in there.
Three things follow, and they are why this post exists.
What lives out there has its own post. What matters here is the one variable the dry route hands to the producer: the water in the fruit, and how fast it leaves.
For the enthusiast: why the container changes the chemistry
In a tank the whole mass shares one temperature and one oxygen level, and the products of the fermentation dissolve away into the water. In a cherry the skin stays intact and the pulp, the mucilage and the seed are sealed in together, so nothing is rinsed away.
The slow reactions include esterification, the joining of an acid and an alcohol that makes most fruit smells. And "too dry for anything to live on" means the water activity, the share of the water that microbes can actually use, has fallen below what the organisms need.
Every gram of pulp and mucilage stays with the seed for the entire dry. There is a great deal of food sitting on the outside of that seed, and three weeks in which something unwelcome can find it.
04 the dry
Three weeks of moving fruit around in the sun.
Water has to leave fast enough that nothing else moves in, and evenly enough that the outside does not seal around a wet middle.
The beans inside a natural start at around 68% water4, and every route finishes at about 10 to 12%78. So the dry route has to take nearly all of the water out of a whole piece of fruit, outdoors, with the sun, without letting anything eat it on the way. Everything below is in service of that one sentence.
What it dries on.
Bare ground is the cheapest and the worst: no air moves under the fruit, and it sits on soil. A patio of concrete or brick is the workhorse: it drains, it holds heat, and it can be swept. Raised beds are fine mesh on a frame above the ground, and air moves under the fruit as well as over it, so a layer dries from both faces: faster, and more evenly.
Thin, turned, covered.
The layer is kept under 40 millimetres deep and raked often12, because a deep layer dries on top and keeps fermenting underneath. It is covered at night, against dew, and in rain. Where a producer cannot cover a lot fast enough, that is the lot that fails.
How long, and what decides.
Whole cherries take roughly 12 to 21 days on a patio12. The spread inside that range is weather. A run of clear dry afternoons and a natural is a three-week job with a predictable end. A wet week in the middle and the fruit sits in the range where things grow, and the lot is in trouble from which it cannot really recover.
For the enthusiast: the starting water, the meter, raking, and dryers
One review puts harvested coffee at 55 to 60% water at the start12; the 68% above is the beans inside naturally processed cherries, measured when the ferment begins4. A producer follows a natural with a moisture meter, morning and afternoon, for the whole of the dry. The only thing that matters in the readings is whether they are still falling.
The same review gives the bed depth and says the cherries should be raked frequently precisely to prevent uncontrolled fermentation and discolouration12. Raked, because every fruit needs its turn on the surface, and because moving the layer breaks up the crust that forms on it. In the first days, when the fruit is wettest and the risk is highest, that means going over the whole bed again and again through the warm part of the day, and covering it at night because dew puts water back into fruit that spent all day losing it. On bare ground, rain turns the ground itself into a reservoir.
On a patio, washed parchment dries in 7 to 15 days against the 12 to 21 for whole cherries12. Mechanical dryers exist and are used, particularly to finish a lot that the weather will not finish. They are not a free pass: the review limits the coffee's own temperature to 45 degrees for cherries and 40 degrees for parchment12, above which you are cooking the seed rather than drying it.
After the dry, the husk rests for a few weeks, which is the most stable form the coffee will ever be in. At the mill the skin, the dried pulp, the mucilage and the parchment all come away in one pass, as a single by-product.
Down in that 10 to 12% corridor the fruit is a hard, dark, rattling husk and the dry is over. It rests a few weeks, then goes to the mill, where everything outside the seed comes away in one pass.
05 the hard one
A bad natural is worse than a bad washed, and it is easier to get one.
Every advantage of the method is also the mechanism by which it fails.
I want to be plain about this, because the romance around natural process skips it. Washing is, among other things, risk control: it takes the food away at the start and ends the fermentation in a tank somebody is watching. A natural gives up both on purpose.
There is a legal floor under all of this. The European Union caps ochratoxin A at 3 micrograms per kilogram in roasted and ground coffee910. Roasting reduces it but does not remove it, which is why the drying bed, not the roastery, is where this gets decided.
For the enthusiast: the acids, the moulds, and the lab thresholds
Over-fermentation. The acids are butyric and propionic. A 2025 trial found that sealing whole cherries under carbon dioxide slowed the loss of their sugars compared with leaving them in air; its authors note that sugar broken down fast in air is what leads to those acids, and that the sealed route favours alcoholic and lactic fermentation7.
Mould. Aspergillus was only 3% of the fungal isolates in the Brazilian survey3, which sounds reassuring until you remember how many isolates that is per cherry and what a few of them can make.
Ochratoxin A. OTA is made on coffee chiefly by Aspergillus ochraceus and A. carbonarius. Bucheli and Taniwaki identified cherry drying, which is to say natural processing, as one of the steps where it can form, because cherries contain sufficient amounts of water to support mould growth and OTA formation during the initial 3 to 5 days of drying5.
The threshold. In vitro, roughly 0.80 water activity, about 14% moisture, is the floor for OTA production on green coffee5; Pardo and colleagues found A. ochraceus grew fastest at 30 degrees and 0.95 to 0.99 water activity and produced no detectable OTA at 0.806.
Uneven drying. Drying from the outside and sealing water in is called case hardening. Part of the lot is still fermenting while the rest is finished.
Labour. A washed coffee spends a day in a tank and then dries as a thin layer of parchment. A natural is levelled, raked and turned repeatedly, and covered when it rains and when the dew comes down.
The law. The EU limit is 5 micrograms per kilogram in instant coffee910. There is no limit for green coffee, and the trade's own guidance is to keep green below about 12.5% moisture in storage10.
Put together: a natural has a wider spread of outcomes than a washed coffee. The best naturals are extraordinary and the washed process cannot make them. The bad ones are undrinkable in a way a bad washed coffee rarely is, and there are more of them. That is not an argument against naturals. It is what you are paying for when one is good.
06 what it tastes like
Sweet, heavy, and carrying the fruit it dried inside.
A natural does not taste like a coffee with fruit added. It tastes like a seed that spent three weeks inside one.
The shorthand: a natural is sweeter and heavier than the same coffee washed, lower in acidity, with fruit notes partly from the plant and partly from the ferment. What it does not tell you is which fruit.
Brazilian naturals are low acid, nutty, chocolate and heavy: most of the world's espresso base. Ethiopian naturals go into berry and stone fruit, sometimes so far that people who have only drunk washed coffee do not believe it is coffee. A process is an amplifier, and what it amplifies is whatever the variety, the altitude and the ripeness of the picking already gave it. A natural made from underripe cherries is not fruity; it is just muddy.
Consistency is a washed-process virtue. A natural buys you a ceiling and charges you a floor for it.
A processing word is not a flavour that was added and it is not a style that was chosen. It is a risk that somebody took, and a set of weeks that somebody worked.coffee/natural-process
One thing I still want to add: a washed and a natural from one farm and one harvest, cupped side by side, because everything above is other people's measurements.
For the enthusiast: floor, ceiling, and how to read the word on the bag
There is a second thing the dry route does that is easy to miss, and it is not a flavour at all. It raises the floor of the sweetness and the body while widening everything else. Two lots off the same farm, processed the same way, in the same three weeks, can land a long way apart, and the reason is in section 05 rather than here.
I have stopped reading the process word on a bag as a flavour promise and started reading it as a statement about what the producer decided to expose the coffee to. Natural says: I left the fruit on, and then I spent three weeks making sure nothing went wrong. That is a claim about labour and about weather before it is a claim about taste, and it is checkable in the cup: a natural that is clean as well as fruity is the evidence that the three weeks went the way they were supposed to. One honest pair of cups of my own would be worth a paragraph.
07 use your agent
Use your agent with this page.
Any part of this page can become a question for your AI.
Every section and figure has an ask agent button. Pick how deep to go (Explain it simply, Go deeper or How do we know?) and copy that prompt. Selecting a passage offers one too. This one covers the whole post:
For the enthusiast: the tools
Each ask agent button opens a menu: pick how deep to go (explain it simply, go deeper, or how do we know), and it copies that prompt. Section 00 offers two more for the whole post. A browser agent that speaks WebMCP calls the page's read-only tools directly. One without a browser reads the plain-text version, natural-process.txt, where every section and figure is numbered.
- start_guided_reading a teaching plan: learn step by step, a quiz, or just the answer.
- get_outline the post's claim, each section's claim, and the figures under their sections.
- explain_section any section for a
child, astudentor anexpert. - explain_figure any figure at the same three levels: what it asks, how to read it, what it shows.
- get_evidence the study behind a claim and what it actually measured.
- get_glossary any technical word on the page, in one line.
I aimed the page at one reading level, and one reading level is a compromise. The three levels are written by hand, not generated at runtime. The tools read this page and nothing else: no network, no key, no backend. Their answers load only when an agent first asks.
Checking whether this browser exposes the tool API.
Every number on this page traces to one of them. The list is also in natural-process.txt.