Urbit / Docs


Urbit is renowned for its exotic terminology. Here's a simple overview from the strange words within.

As Dijkstra put it: “The purpose of abstraction is not to be vague, but to create a new semantic level in which one can be absolutely precise.”


Also known as an “app,” an application is a Hoon program that can hold state.


An arm is a named, functionally-computed attribute of a core.

The hoon of each arm is compiled to a Nock formula, with the enclosing core itself as the subject.

There are two kinds of arms: dry and wet. Most arms are dry.

  • dry:

    normal, ++, polymorphic by means of variance

    For a dry arm, we ask: is the new payload compatible with the old payload (against which the core was compiled)?

  • wet:

    unusual, +-, polymorphic by means of genericity

    A wet arm uses the hoon as a macro. We create a new type analysis path, which works as if we expanded the callee with the caller's context.

See advanced types.


The Urbit operating system and kernel. Arvo's state is a pure function of its event log, and it serves as the Urbit event manager. It contains vanes, which are kernel modules that perform essential system operations. Arvo itself can only deal with one file at a time. For more, it relies on Ford, the build-system vane.

Arvo being purely functional means that the it is a function that is:

(state, event) → (state, effects)

Arvo coordinates and reloads vanes. It can be thought of as a traffic-director. Any vane needs to go through Arvo before reaching another vane. Events and their effects happen like so:

Unix event → vere (the virtual machine that runs Urbit) → Arvo → vane → Arvo

Arvo is located in /home/sys/arvo.hoon within your urbit.

More in-depth information on Arvo can be found here.

Below are Arvo modules, which are called “vanes”.

  • Ames

    The networking vane. Ames handles all things UDP. Here live the protocols for discovering and interacting with other urbits. It’s also the name for the Urbit network itself. All communication is signed and encrypted.

    Ames is located in /home/sys/vane/ames.hoon within your urbit.

    More in-depth information on Ames can be found here.

  • Behn

    The timing vane. Behn allows for applications to schedule events, which are managed in a simple priority queue. For example, Clay, the Urbit filesystem, uses Behn to keep track of time-specific file requests. Eyre, the Urbit web vane, uses Behn for timing-out HTTP sessions.

    Behn is located in /home/sys/vane/behn.hoon within your urbit.

  • Clay

    The filesystem and typed revision-control vane. Think of Clay as a continuously synced git. It handles file-change events and maps them from Urbit to Unix and vice versa.

    A common way to use it is to create a pier, a directory that exists in and is visible in Unix. Changes are automatically recorded from Urbit to the Unix directory, and vice versa. Just set it and forget it!

    Clay is located at /home/sys/vane/clay.hoon within your urbit.

    Basic information on using Clay can be found here. An in-depth explanation of the architecture can be found here.

  • Dill

    The terminal-driver vane. You run your urbit in your Unix terminal, and Unix sends every event -- such as a keystroke or a change in the dimensions of the terminal window -- to be handled by Dill. Dill acts as an intermediary for anything that uses keyboard events. This is why there’s some input lag in Urbit’s command-line interface. (Optimization will come later!)

    A keyboard event's journey from Unix to Dojo, the Urbit shell, can be imagined as diagrammed below:

    Keystroke in Unix -> Vere (virtual machine) -> Arvo -> Dill -> the Dojo

    Dill is located at /home/sys/vane/clay.hoon within your urbit.

    More in-depth information on Dill can be found here.

  • Eyre

    The web-server vane. Anything HTTP-related lives here. Unix sends HTTP messages though to Eyre, and Eyre produces HTTP messages in response.

    In general, apps and vanes do not call Eyre; rather, Eyre calls apps and vanes. Eyre uses Ford and Gall to functionally publish pages and facilitate communication with apps.

    Eyre is located at /home/sys/vane/eyre.hoon within your urbit.

    More in-depth information on Eyre can be found here.

  • Ford

    The build-system vane. Ford is capable of sequencing generic asynchronous computations. Its uses include linking source files into programs, assembling live-updating websites, and performing file-type conversions.

    Ford neither accepts Unix events nor produces Unix effects. It exists entirely for the benefit of Urbit applications and other vanes, in particular Gall.

    Ford is located at /home/sys/vane/ford.hoon within your urbit.

    More in-depth information on Ford can be found here.

  • Gall

    The application-management vane. Userspace apps -- daemons, really -- are started, stopped, and sandboxed by Gall. Gall provides developers with a consistent interface for hooking their app up to the operating system. It allows applications and other vanes to send messages to applications and subscribe to data streams. Messages coming into Gall are routed to the intended application, and the response comes back along the same route. If the intended target is on another ship, Gall will route it, behind the scenes, through Ames to the other ship.

    Gall is located at /home/sys/vane/gall.hoon within your urbit.

    More in-depth information on Ford can be found here.


An atom is any natural number, including zero. The atom is the most basic data type in Nock and Hoon. A Hoon atom type describes a Nock atom with two additional pieces of metadata: an aura, and an optional constant.

An atom type is warm or cold based on whether the constant exists.

  • warm:

    If the constant is ~ (null), any atom is in the type.

  • cold:

    If the constant is [~ atom], its only legal value is atom.

See basic types


An aura is a soft atom type. They appear as strings beginning with @. Auras represent the structure of an atom, print format, or other semantics. Its constraints on the value of an atom aren't enforced in any way.

You can cast something to an aura, like a decimal, and get back its hexademical. For example:

    > =a `@`29
    > ^-  @ux  a

Auras are an advisory type system. You can go up or down, but not sideways in the hierarchy. @t and @ can each be cast to the other, but @t can’t be cast to @u, for example, without first being cast to the empty aura @.

For example

    > ^-  @  'c'                       ::  'c' is of the aura @t

    > ^-  @u  'c'
    ! exit

Here is a non-exhaustive list of auras:

Aura         Meaning                        Example Literal Syntax
@d           date
  @da        absolute date                  ~2018.5.14..22.31.46..1435
@n           nil                            ~
@p           phonemic string (ship name)    ~sorreg-namtyv
@r           IEEE floating-point
  @rd        double precision  (64 bits)    .~6.02214085774e23
  @rh        half precision (16 bits)       .~~3.14
@s           signed integer, sign-bit low
  @sb        signed binary                  --0b11.1000
  @sd        signed decimal                 --1.000.056
  @sx        signed hexadecimal             -0x5f5.e138
@t           UTF-8 text (cord)              'howdy'
  @ta        ASCII text (knot)              ~.howdy
    @tas     ASCII text symbol (term)       %howdy
@u              unsigned integer
  @ub           unsigned binary             0b11.1000
  @ud           unsigned decimal            1.000.056
  @ux           unsigned hexadecimal        0x5f5.e138

You can find the list of all auras here.


Blocksize is the power-of-two bitwidth of an atom. So, an atom with a blocksize of 3 has a bitwidth of 2^3, which is 8. A bitwidth of 8, in turn, can represent 2^8 values, or 256.


To cast is to put a Hoon expression through the compiler’s type-checker to test whether the expression is guaranteed to evaluate to a product that falls within the desired type. If the expression is not guaranteed to behave in this way, then the compiler will halt with a nest-fail crash. Proper Hoon code will include a cast for the product type of each gate.

The cast rune is ^- (pronounced “ket-hep”). So, to cast to a list of atoms, we write ^- (list @). The irregular form of that expression is (list @).


In Nock and Hoon, a cell is an ordered pair of nouns.

Syntax: [a b].

Cells orient to the right; for example, the noun [2 6 14 15] is short for [2 [6 [14 15]]].


A clam is the use of a structure as a gate. Usually used to verify marks from other urbits.

    > ((list @t) [97 98 ~])
    <|a b|>


In Hoon, cord is an aura for representing atoms as strings. UTF-8, least-significant bit first. Represented as the aura @t.

    > `@t`478.560.413.032


A core is a cell of [code data], where we call the code head the battery and the data tail the payload. All code-data structures in normal languages (functions, objects, modules, etc.) translate to cores in Hoon.

  • battery

    The code-containing head of a core.

  • payload

    The data-containing tail of a core.

See basic types.


The Dojo is the Urbit shell; the command line where you run programs, expressions, and system commands. It’s a CLI that is encountered right when an urbit is started for the first time. It is somewhat analogous to Bash, the Unix terminal, and more analogous to a Hoon REPL.

You’ll notice that many things can’t be written here; this is normal. The Dojo enforces syntactical correctness right at the command line. Strings that can’t be continued into a valid expression are automatically trimmed back to something that can be. Some syntactically invalid expressions can be completed, but the Dojo will prevent them from being entered as a command. You can play around with this functionality learn what syntax is kosher.


A duct is an Arvo-level call stack. At the beginning of every duct is a Unix event, such as a keystroke, network packet, timer event, or file change.

A duct is a list of wires. A wire is itself a list of knots. A knot is, in turn, a knot.


A face is the Hoon analog to what is a variable name in most other programming languages. Faces are metadata labels for legs, allowing us to address parts. Because there is no metadata in Nock, faces cannot exist in Nock.

    > a=2

In the above example, a is the face and 2 is the value.

Hoon has no scope or symbol-table; there is only the subject. To “declare” a “variable” is to construct a new subject: [name=value old-subject].

See advanced types.


A flag is Hoon's version of a boolean. 0 (%.y) is yes, 1 (%.n) is no.

Why? It's fresh, it's different, it's new. And it's annoying. And it keeps you on your toes. And it's also just intuitively right.


A gate is a core with one arm -- Hoon's closest analog to a function. To call a gate on an argument, replace the sample (at tree address +6 in the core) with the argument, and then compute the arm.

The payload of a gate has a shape of [sample context].

  • sample

    The argument tuple.

  • context

    The subject in which the gate was defined.

See basic types, %- (“cenhep”) (the rune for calling a gate).


A generator is like a limited Hoon script: it’s a shell command that produces a value before disappearing. Generators take an input noun (a value) and produce an output noun by performing calculations on its arguments, prompting for input, or making HTTP requests.

You need a bar rune like |= in your generator, because a generator is essentially a function that you’re calling.

Generators, unlike apps, hold no state. They are pure, stateless functions, and they cannot send moves. Because of this, generators are have a more narrow use than apps. Among the non-trivial uses of a a generator is moving data into apps -- they’re the backbone of Urbit’s information pipeline.

Generators are run from the /gen directory of your current desk. For example, to run a generator named whatever.hoon, you would type > +whatever in the Dojo.


A glyph is a single, non-alphanumeric ASCII character. Each glyph has its own monosyllabic name, which allows the efficient communication of Hoon-related information in speech. Runes are composed of two glyphs. Below is a complete list of glyphs with their associated pronunciations.

ace  [1 space]       hep  -            rob  }
ban  >               ket  ^            say  '
bar  |               lac  [            sig  ~
bat  \               led  <            tar  *
bus  $               lit  (            tec  `
cab  _               lob  {            tis  =
cen  %               lus  +            vat  @
col  :               mic  ;            wut  ?
com  ,               net  /            yel  "
dot  .               pad  &            zap  !
gap  [>1 space, nl]  rac  ]
hax  #               rit  )


Hall is the Urbit communication back-end. It’s a messaging bus that can work with clients of all sorts. :talk, the built-in chat client, is one such app that uses Hall.

More information on Hall can be found here.

  • circle

    A circle is the “audience” of your messages -- who can receive them. It can be thought of as a chat channel, but this is a little too narrow. More specifically, it’s a named collection of messages created by and hosted on a ship’s Hall, usually represented as ~ship-name/circle-name. On Talk, Urbit’s built-in chat front-end, the first circle that you will likely encounter is /urbit-help.


Hood orchestrates many of the Urbit initialization systems necessary for boot.


Hoon is a strict, higher-order typed functional language that compiles itself to Nock.

The Hoon source file is located in /home/sys/hoon.hoon within your urbit.

More information can be found in the hoon section.

  • Mint

    Mint is the Hoon compiler function. Mint takes the subject type and the expression source (a hoon, for example) and produces the product type and nock formula. So [type hoon] is mapped to [type nock].


A hoon (lowercase) is the result of parsing a Hoon source expression into an AST node. These AST nodes are nouns, like all other Hoon data. Because every Hoon program is, in its entirety, a single expression of Hoon, the result of parsing the whole thing into an AST is a single hoon.

A hoon is a tagged union of the form [%tag data], where the tag is a constant such as %brts (from the source rune |=, i.e. “bartis”), and is matched up with the appropriate type of data (often more hoons, from source subexpressions). For example, the expression :-(2 17), once parsed into an AST, becomes the following:

[%clhp p=[%sand p=%ud q=2] q=[%sand p=%ud q=17]]

The %clhp is produced from the rune :- (i.e. “colhep”). The 2 and 17 have each been parsed as %sand-tagged hoons, which represent atoms (in this case each with an aura of %ud, i.e. unsigned decimal).

To parse Hoon source into a hoon AST, use ream on a cord containing Hoon source. Try the following in Dojo:

(ream ':+(12 7 %a)')

The result should be:

[%clls p=[%sand p=%ud q=12] q=[%sand p=%ud q=7] r=[%rock p=%tas q=97]]

See expression reference for the full list of hoons.

irregular form

Irregular forms are alternative ways of writing Hoon expressions, alternatives to the digraphic runes such as :~. Irregular forms are usually more compact than regular forms.=

See irregular forms.


A jet is a piece of C code that have equivalent products to a piece of Hoon or Nock code. They are used for performance reasons; jets are meant to be faster than the code that they are substituting for.

Jets are necessary for mathematics, performance-wise, because Nock handles it so inefficiently. Nock, for example, can’t decrement (reduce a number by one) except by starting building a loop that starts from zero and increments to a number just below.

See API overview.


A lark is an expression that returns a limb at the specified address in a cell. Using - by itself returns the head of the subject, and using + by itself returns the tail:

> -:[[4 5] [6 [14 15]]]
[4 5]

> +:[[4 5] [6 [14 15]]]
[6 14 15]

You can combine + or - with either of > or < to get a more specific limb of the subject. -< returns the head of the head; -> returns the tail of the head; +< returns the head of the tail; and +> returns the tail of the tail. By alternating the +/- symbols with </> symbols, you can go as deep as you desire.

> +>-:[[4 5] [6 [14 15]]]

> +>+:[[4 5] [6 [14 15]]]


A limb is an attribute or variable reference. A limb is an arm or a leg.

To resolve a limb named “foo”, the subject is searched depth-first, head-first for either a face named “foo” or a core with an arm of “foo”. If a face is found, the result is a leg, if a core is found, the result is the product of the arm.

  • leg

    a subexpression, or subtree of the subject.

  • wing

    a search path into the subject, composed of limbs. Search occurs from right to left (a.b means b within a).

See Limbs and wings


A move is the Arvo equivalent of a syscall.


In Nock and Hoon, a noun is an atom or a cell.


Nock is a Turing-complete, non-lambda combinator interpreter. It's Urbit's low-level programming language. Nock is functional and typeless.

  • noun
    an _atom_ or a _cell_.
  • atom

    any natural number, including zero.

    • cell

      any ordered pair of nouns.

    • cell

      any ordered pair of nouns.

    • subject

      a noun - the data against which a formula is evaluated.

    • formula

      a noun - a function at the nock level.

    • product

      a noun - the result of evaluating a formula against a subject.

See the Nock definition.


A mark is Urbit's version of a MIME type, if a MIME type was an executable specification. The mark is just a label that's used as a path to a local source file in the Arvo filesystem. This source file defines a core that can mold untrusted data, diff and patch, convert to other marks, etc.

If this sounds like magic, it isn't quite magic. There's no way for different urbits to make sure they mean the same thing by the same mark. However, when incompatibility happens, marks ensure that we at least handle the situation in a predictable way.


Every core has a metal which defines its variance model (ie, the properties of the type of a compatible core). The default is gold (invariant).

  • gold: invariant

  • lead: bivariant

  • zinc: covariant

  • iron: contravariant

See advanced types.


An Urbit ship is: a cryptographic title on a will signed by private key; a human-memorable name; and a packet-routing address. Ships are classed by the number of bits in their address:

    Size   Name    Parent  Object      Example
    -----  ------  ------  ------      -------
    2^8    galaxy  ~       supernode   ~zod
    2^16   star    galaxy  supernode   ~dozbud
    2^32   planet  star    user        ~tasfyn-partyv
    2^64   moon    planet  device      ~sigsam-nimbot-tasfyn-partyv
    2^128  comet   ~       bot         ~racmus-mollen-fallyt-linpex--watres-sibbur-modlux-rinmex

Any ship can be called an urbit. The lowercase “u” distinguishes it from Urbit the software stack and Urbit the network.

You can find a longer-form summary here..

An Urbit identity is a string like ~firbyr-napbes. It means nothing, but it's easy to remember and say out loud. ~firbyr-napbes is actually just a 32-bit number (3.237.967.392, to be exact), like an IP address, that we turn into a human-memorable string. The full name of this string can be viewed by typing our in the Dojo, Urbit's shell. This is useful when running a ship with a longer name, such as a moon or a comet.

Technically, an urbit is a secure digital identity that you own and control with a cryptographic key, like a Bitcoin wallet. As in Bitcoin, the supply of urbits is mathematically limited. This keeps the network friendly, by making spam and abuse expensive.

An urbit name is just a number; smaller numbers make shorter names. Shorter names are easier to remember, so they're more valuable. So urbits are classified by the number of bits in their name. (A ship name is just a scrambled base-256 representation of the number.)

A 32-bit urbit (like ~firbyr-napbes) is called a planet. A 16-bit urbit (like ~pollev) is a star. An 8-bit ship (like ~mun) is a galaxy. A planet is an identity for an independent, adult human. Stars and galaxies are network infrastructure.

Each planet or star is launched by its parent, the star or galaxy whose number is its bottom half. So the planet ~firbyr-napbes, 0xdead.beef or 3.735.928.559, is the child of ~pollev, 0xbeef or 48.879, whose parent is ~mun, 0xef, 239. The galaxy at the address of 0 is ~zod.

  • pier

    A ship's pier is its Unix directory. For planets, the name of the pier is usually the planet name.


A structure is an idempotent gate (function) that constructs and validates types in Hoon.

An example: Arvo marks use Hoon's type system via structures to validate untrusted network data.

Here's some common structure terminology:

  • bunt:

    A bunt is the default value of a structure. It's also a verb; to bunt a structure is to produce the bunt of that structure. For example, bunting @ results in 0, and bunting @t results in ''. Bunting is performed with * followed by the structure in question:

        > *@
        > *@t
  • icon: The type of the mold's range.

See mold hoons.


A Hoon rune is a pair of ASCII symbols used to begin a Hoon expression.

For example, the rune ?: is Hoon's most common conditional, a branch on a boolean test. The first symbol in a rune represents a family of related runes. For example, the ? family are all conditionals.

The result of parsing a Hoon source expression&mdash;the rune, followed by its respective children&mdash;into an AST node is simply called a hoon.

Runes have two syntactic forms, tall and flat:

  • tall:

    multiple lines, no parentheses, two or more spaces between tokens


    > %+  add  2  2
  • flat:

    one line, parentheses, one space between tokens


    > %+(add 2 2)                              ::  regular flat form
    > (add 2 2)                                ::  irregular flat form

Tall hoons can contain flat hoons, but not vice versa. All irregular forms are flat.

_See hoon concept, expressions.


Sail is the Hoon markup language for XML.


To scry means to request data from the Arvo namespace and bring it to userspace.

See .^ (dot-ket).


slot is the name for Hoon's tree-addressing scheme.

Every cell has a head and a tail, each of which may be either an atom or a cell. Therefore every noun is a binary tree. +1 or . resolves to the whole cell (technically this would work against an atom as well). The head of +n is +2n, the tail is +(2n+1).

+7 is a special address for gates, because the position is defined (by convention) as the context of a gate and of all cores.


Talk is Urbit's built-in chat app. It’s one example of an app that can be built on top of Hall, the Urbit back-end messaging system.

See Messaging. for more information.


A Hoon type defines a set (finite or infinite) of nouns and ascribes some semantics to it. There is no direct syntax for defining types; they are always defined by inference (i.e., by mint), usually using a constructor (structure).

All types are assembled out of base types defined in ++type. (Look up ++ type in hoon.hoon for examples.) When the compiler does type-inference on a program, it assembles complex types out of the simpler built-in types.

  • nest

    nest is an internal Hoon function on two types which performs a type compatibility test. nest produces yes if the set of nouns in the second type is provably a subset of the first. If nest produces no, the Hoon programmer will receive a nest-fail error. This is one of the most commons errors in Hoon programming.

    See advanced types. See troubleshooting.


Hoon's equivalent of a nullable pointer or a Haskell Maybe. If q is ~, null, the type is warm; any atom is in the type. If q is [~ x], where x is any atom, the type is cold; its only legal value is the constant x.


A vane is an Arvo module. Each Vane is a single file because Arvo cannot, on its own, deal with programs that are split between multiple files. Some vanes can do this, however.

Vere (pronounced “vair”)

The runtime system, written in C, that implements Nock. All It’s written in C. Vere-level errors are prepended with as bail:.

Vere is a virtual machine: it pretends to be hardware, and is the single point of contact between your Urbit and the rest of your system. But more than that, Vere handles the Urbit event log and handles jets. It’s also contains the i/o drivers for Urbit’s vanes, responsible for generating events from Unix and applying effects to Unix.


Zuse is a defined interface for vanes. It contains the part of the Hoon standard library that is specific to Arvo functionality.

Arvo is located in /home/sys/zuse.hoon within your urbit.