ActuatorFeature flags + dynamic config + segments

One service, one rule language, evaluated against rich context. Fast SDKs, full audit trail, designed so AI agents and humans can operate it together.

Get started

API reference

Concepts

Flags, configs, and segments together

Boolean and string feature flags. Typed dynamic configs (string, number, boolean, JSON). Reusable named segments composed inside rules. One ruleset, one mental model.

Rich context, fast evaluation

Rules match against any user attribute you pass — plan, role, country, region, signup date. RE2 regex, percentage rollouts hashed for stickiness, depth-bounded boolean trees.

Built for agents

Every evaluation returns a structured reason. A propose / apply workflow lets an agent stage a change, present the blast radius to a human, and only mutate after approval. First-party MCP server.

Full audit trail

Every change records actor, delegator, approver, before, after, diff, and reason. Cursor-paginated audit query API, time-travel evaluation with `asOf`, history endpoint that replays changes against any context.

What's in here

Getting started — sign up, create your first flag, evaluate it from an SDK.
Concepts — what flags, configs, segments, rules, and contexts are, and how the evaluator combines them.
API reference — every endpoint, request shape, response shape, error code.
Agents and AI — scoped tokens, the propose / apply flow, counterfactual preview, the MCP server.

SDKs

TypeScript — @actuator/sdk. Used by the dashboard; also installable into any Node or browser project.
Swift — for iOS, macOS, and server-side Swift.
MCP — @actuator/mcp. A Model Context Protocol server exposing the API as agent tools. Drop into Claude Desktop, Claude Code, or any MCP host.

If you'd rather skip an SDK, every operation is a plain HTTPS call against a documented JSON contract.