AI Rule Engine Docs
Go to App

Glossary

Definitions for the terms used throughout this documentation.

TermDefinition
Access policyOrganization-level controls over sign-in and sessions — multi-factor authentication, idle and session timeouts, credential lifetime caps — plus an Enterprise IP allowlist.
Access reviewA point-in-time report of every member’s access to every environment in the organization, including access granted implicitly through organization-admin status.
ActionA step that runs when a rule's condition matches — for example calling an API, transforming data, writing to context, logging, or running an AI prompt.
Action setA named, reusable list of actions that a rule can reference instead of defining actions inline.
ActivationA single rule firing during a forward-chaining run. The Max activations setting caps how many may occur before the run halts.
Active date windowThe optional effective/expiration dates on a rule. Outside the window the rule is skipped exactly like a disabled rule — its condition is never evaluated and its actions never run.
Audit logThe append-only, newest-first record of every change to rules, releases, access, and credentials in an organization, filterable and (on Enterprise) exportable or streamed to a SIEM.
AuthoringDesigning your logic — organizations, projects, rule sets, conditions, and actions — as opposed to running it.
ClauseThe smallest part of a condition: a context value, an operator, and a value to compare against.
Clause groupA set of clauses (and nested groups) combined with All, Any, or None logic.
Compliance archiveAn Enterprise add-on that writes a daily, write-once copy of the audit log and release evidence to immutable storage you own, kept for seven years.
ConditionThe test that decides whether a rule matches. Built from clauses and groups; can be saved and reused or defined inline.
ContextThe working data of a single run: the input you provide plus anything actions add as they execute.
Context keyThe address used to read or write a value in the context, such as ticket.subject.
Correlation IDAn identifier assigned to each run that ties together its input, files, and log entries.
Coverage gapA combination of input values that no rule in a rule set matches, reported by verification and shown in the decision table as a “Not covered” row.
CreditPrepaid balance used for metered usage such as runs beyond your plan's allowance and AI generation.
Custom roleAn organization-defined, additive role — up to 25 per organization on Enterprise — that grants extra permissions on top of the built-in Read, Write, Run, Debug, and Admin roles.
Data type / Data valueA value's type (String, Number, Boolean, DateTime, Blob, or Url) and the typed value itself.
Dedicated hostingA hosting option where an environment runs in your own cloud subscription.
Deployment pipelineAn org-scoped, ordered list of stages, each bound to one environment with its own approval mode, that a project release is promoted through.
Derived factA context value computed on demand from other values the first time a rule reads it, rather than supplied as run input.
Disaster recoveryAn Enterprise feature that backs up an entire organization — projects, conditions, action sets, rule sets, and tests — nightly to storage you own, with on-demand backup, download, and restore.
Effective dateThe inclusive start of a rule's active date window, in UTC. Before it, the rule is not evaluated. Blank means the rule has no start bound.
EnvironmentAn isolated place where rule sets run, owned by an organization, with its own settings, variables, schedules, storage, and logs.
ExecutingRunning a rule set inside an environment, as opposed to authoring it.
ExpressionA small formula that computes a value from other values, such as ctx.price * ctx.qty. Usable wherever a value is read, and as the left side of a clause.
Expiration dateThe exclusive end of a rule's active date window, in UTC. At and after it, the rule is not evaluated. Blank means the rule never expires.
ExtensionA reusable, versioned unit of functionality that a rule can invoke through the Extension action.
FindingOne issue reported by verification, carrying a severity of Error, Warning, or Info and naming the rule, context key, or input region it concerns.
Backward chainingThe goal-driven inference strategy: the engine starts from the keys you want established and fires only the rules needed to produce them. Also called goal-driven evaluation.
Forward chainingThe inference strategy where rules re-fire as other rules change the context they read, until the run reaches a stable state.
GoalA context key a goal-driven run works backwards from. Declared on the rule set, or overridden for one run.
Free tierA no-cost hosting option on shared infrastructure with reduced limits.
Governance policyOrganization-wide release controls (Team plan and above) that set a floor every project and pipeline must meet or exceed — requiring approval, requiring tests to pass, requiring an approval stage, and enforcing four-eyes on promotions.
Hosting optionWhere an environment runs: Dedicated, Shared, or Free tier.
Human interventionA pause in a run while a person reviews and approves before it continues.
InferenceThe opt-in execution mode that turns on forward chaining for a rule set, firing rules one at a time by salience and re-evaluating as the context changes.
MCPModel Context Protocol — lets an AI assistant connect to your account so it can run your rule sets and read files and logs.
Machine API keyAn organization-admin-issued credential with granular permissions, environment scoping, and an optional lifetime cap — distinct from the project/rule-set-scoped keys on the API Access screen.
OAuth & app authorizationsHow you authorize third-party apps to act on your behalf, and review or revoke that access.
OrganizationThe ownership and billing boundary that holds members, plan, credits, projects, and environments.
Plan / TierThe subscription level (Free, Pro, Team, or Enterprise) that sets an organization's limits and price.
ProjectA container for related rule sets, conditions, and action sets within an organization.
Project releaseAn immutable, sequentially numbered snapshot naming the exact versions of every ruleset, condition, and action set in a project at one moment, promoted through a deployment pipeline.
RuleA single unit of logic: one condition plus one or more actions that run when the condition matches.
Rule setAn ordered collection of rules plus a context definition; the unit you run.
SalienceA per-rule priority used only in inference: the highest-salience eligible rule fires first, with ties falling back to rule order. In goal-driven runs it decides which rule is preferred to establish a key.
Shared hostingA hosting option where environments run on shared infrastructure managed for you.
TemplateA prebuilt project or rule set you can import as a starting point.
Trace (inference)A firing-by-firing record of an inference run — every firing, why each condition passed or failed, and each context change — viewable in run history and explainable with AI.
VariableA named value stored in an environment that rules and actions can read while running.
VerificationA static check of a rule set that reads it instead of running it, reporting rules that can never fire, duplicates, conflicting writes, broken references, and coverage gaps.
What-if comparisonA sandboxed re-run that changes selected inputs and shows which outcomes flip versus a baseline, with a provenance chain back to the inputs.
See also New to these terms? Start with Core Concepts and the Getting Started walkthrough.