B Language on one page

The whole language of tok 1.0 on one page, as tok cheatsheet prints it and as the skill for AI agents carries it. The normative description is the specification, docs/spec.md.

One statement per line. Blocks by indentation (spaces). # starts a comment. Every statement starts with a keyword. Identifiers: [A-Za-z_][A-Za-z0-9_-]*, unique. Labels: “quoted text”. Attributes: key=value (value: word, number or “text”) and bare flags. A key or flag is written once per line, the triggers of multiple events aside (in 1.0 a repeated one is the deprecation warning R090).

tok 1.0 "Title" [executable]                       # header, first line (0.1–0.7 documents stay valid)
    [process-name="Name"] (process of a document without pools)   doc="Text" on any statement: documentation

start ID "Label" [message="M" | timer=P3D | signal="S" | condition="feel"] [form=F]
end ID "Label" [message="M" | error="E" | escalation="X" | signal="S" | terminate]   # error/escalation: [code="E1"]
catch ID "Label" message="M" | timer=PT2H | signal="S" | condition="feel" | link="L"
throw ID "Label" [message="M" | escalation="X" | signal="S" | link="L" | compensate]
boundary ID "Label" on=ACTIVITY timer=P7D [noninterrupting]   # also message/error/escalation/signal
    several triggers = multiple event: message="A" message="B" signal="C"; + parallel = all of them (catching)
    trigger not given yet: flags message timer signal condition error escalation (start ID "Label" message)

task | user | manual | service | script | rule | send | receive | call   ID "Label"
    flags: loop parallel sequential compensation
    user … form=F        rule … decision=D result=var        script … expr="feel" result=var
    service … type="job"  send/receive … message="M"          call … process=P | task=GLOBAL
    loop while="feel" [test-before] [loop-max=5]              receive … instantiate (starts the process)
    service|send|receive … operation=OP                       user … by=RESOURCE (performer)
sub ID "Label"           # embedded sub-process: indented block with its own start and end
adhoc ID "Label"         # ad-hoc sub-process: activities in any order, no start or end events
eventsub ID "Label"      # event sub-process: block starts with a start event with a trigger
transaction ID "Label" [method=compensate|store|image]   # block like sub; inside: end ID cancel
boundary ID on=TRANSACTION cancel                        # leaves the transaction after cancel

xor | and | or | eventgw | complex   ID ["Question?"]
complex ID "?" activation="_activationCount >= 2"        eventgw ID instantiate [parallel]

pool ID "Label" [process=P] [process-name="Name"] [executable]    # with a block: participant + process
  lane ID "Label"                           # lanes nest; nodes go inside lanes
pool ID "Label" [min=2] [max=5]             # no block: black-box pool; min/max: several participants (0.6)

flow A -> B -> C                            # chain of sequence flows
flow G -> D "tak" when="amount > 100"       # one flow: label, FEEL condition
flow G -> E "nie" default                   # default flow; when="" = condition not written yet
message A -> B "Label" [id=ID]              # message flow between pools (nodes or pools)

conversation ID "Label" pools="P Q" [messages="M1 M2"]       # conversation between pools (message flow ids)
subconversation ID "Label" pools="P Q"                       # block: conversation nodes
callconversation ID "Label" pools="P Q" collaboration=ID [map="inner:outer; …"]

choreography ID "Label"                     # block: start/end/catch/throw/boundary, xor and or eventgw complex, flows
  choreo ID "Label" from=P to=Q [message="M"] [reply="R"] [loop|parallel|sequential]
  subchoreo ID "Label" pools="P Q R" from=P   # block like a choreography
  callchoreo ID "Label" pools="P Q" from=P choreography=ID [map="inner:outer; …"]

data ID "Label" [state="złożony"] [type=ITEM] [object="Data object"] [collection]
store ID "Label" [type=ITEM]
reads ACTIVITY DATA [input=ID]   writes ACTIVITY DATA [output=ID]   # DATA: data, store, input, output, property
    reads END|THROW DATA   writes START|CATCH|BOUNDARY DATA          # data of events
item ID "Label" [structure=Name|xsd:string|p:Name namespace="URI"] [kind=information|physical] [collection]
input ID "Label" [type=ITEM] [collection] [on=ACTIVITY|EVENT]    output ID … (same)
property ID "Label" [type=ITEM] [on=ACTIVITY|EVENT]

global ID "Label" [kind=task|user|manual|script|rule]       # called by call … task=ID
interface ID "Label"
  operation ID "Label" in="Message" [out="Message"] [error="Error" …]
key ID "Label"                               # correlation key (needs pools)
  prop NAME message="Message" path="expression"
note ID "Text" [on=ID]       group ID "Label" members="A B C"

decision ID "Label" hit=first               # unique | first | any | priority | collect
  | in1: string | in2: number || out: string || annotation |
  | "e-book"    | -           || "manager"   || licences   |
  | -           | <= 150      || "auto"      ||            |

# decision requirements graphs and boxed expressions (language 0.7, header tok 0.7 or tok 1.0)
inputdata ID "Label" [name="FEEL name"] type=number            # also date, "date and time", list, an item …
source ID "Label" [requires="D I S"] [kind="regulation"] [uri="https://…"]
bkm ID "Label" params="a: number, b: string" [type=T] [requires="BKM S"] expr="a * b"
decision ID "Label" [name="FEEL name"] [type=T] requires="D I BKM S" [question="?"] [answers="…"] expr="feel"
decision ID "Label" requires="…" invoke=BKM with="a = feel; b = feel"     # invocation
decision ID "Label" requires="…" hit=rule-order|output-order|collect aggregation=sum|count|min|max default="0"
  | in: number = [0..100] || out: string = "high", "low" |              # input/output values
decision ID "Label"                         # context: entries in order, optional result last
  entry NAME [type=T] expr="feel"           # NAME a word or "quoted name"; each line may have its own block
  entry f params="x: number" expr="x * 2"   # function definition
  entry table                               # nested table or relation
    | col: string | val: number |           # relation: header without ||
    | "a"         | 1           |
  result expr="f(2)"
decision ID "Label"                         # list
  element expr="1"
item T "Label" [type=number] [allowed="[0..10]"] [collection]      # DMN type (0.7)
  item field type=string                    # components

FEEL of 0.7 and 1.0 documents is full FEEL (level 3): for x in l return e, some/every x in l satisfies c, l[item > 1], {a: 1}.a, @"2027-01-15", date and time(…), durations, ranges, names with spaces resolved against the document.

form ID "Title"
  text "## Markdown with {{variable}}"
  field KEY "Label" type=textfield|textarea|number|checkbox|select|radio|datetime [required] [options="v:Label, v:Label"]

simulate cases=200 seed=1 arrival="exp(2h)"
resource ID "Label" count=2
variable NAME "unif(20, 300)"               # fixed unif tri norm lnorm exp bern cat; units s m h d
    activity … by=RESOURCE duration="tri(10m, 20m, 1h)" assign="zgoda=bern(0.8)"
    catch … delay="exp(3d)"     flow … p=0.3

FEEL in when, expr, condition, correlation and decision cells is written without a leading =.

Compensation and catch-all events (any document): service a … compensate=h + service h … compensation · throw c … compensate [activity=a] · boundary b on=a error (any error) · boundary b on=a escalation (any escalation)

Executable layer (only with executable; values computed at run time start with =):

user t … assignee="kierownik" candidate-groups="a, b" due="=now() + duration(\"P3D\")" priority=70
user t … candidate-users="ola, jan" follow-up="2027-01-20T09:00:00Z"
user t … parallel collection="items" item=x output=results output-item="x.result" until="numberOfCompletedInstances >= 2"
adhoc a … active="chosen" until="done = true" wait
call c … process=child binding=deployment|latest|version-tag version="v2" child-vars=none parent-vars=none business-id="=numer"
service s … type="job" retries=5 headers="key: value; …" inputs="target = expr; …" outputs="target = expr; …"
service r … connector=rest method=GET url="https://…" result=response [body= query= headers-in=]
end e … message="M" type="job"          catch|boundary … condition="x > 1" events="create, update"
any … listeners="start: job-type; end: job-type (5)"   (user tasks also: creating, assigning, updating, completing, canceling)
tok 0.3 "T" executable version-tag="v1"   pool p … version-tag="v1"

Language 0.5 elements (transaction, cancel, complex, loop conditions, multiple events, instantiate, process data, global tasks, operations, keys) and 0.6 elements (conversations, choreographies, min/max) are BPMN only: --level basic|analytic rejects them, and in executable documents for Camunda they are reported as C100–C123. Decisions of 0.7 are level full; Camunda codes C130 (bkm with invoke=), C131 (non-literal function body), C132 (decision without logic). In 0.5 documents resource and by= also export BPMN performers.

Commands: tok check F [--class descriptive|analytic|common-executable|full] · tok fmt F [-w] · tok bpmn F · tok svg F · tok export F [-o DIR] (0.7: NAME.dmn + NAME.drd.svg per decision graph) · tok import F.bpmn|F.dmn [-o F.tok] (BPMN: writes F.tok.layout, the drawing reused by bpmn/svg/export; --no-sidecar) · tok sim F · tok eval F DECISION name=value … (0.7: with required decisions; JSON values x='{"a": 1}') · --lang pl · --json