# Weir > Weir is a controlled private-alpha native onboarding delivery and experimentation system for > coding-agent workflows. The v4 API, flows, and dashboard are live; compatible SwiftUI and React > Native/Expo builds have qualified dogfood evidence. Public release, reproducible packages, and > complete Android device proof remain gated. > Spec + schema are public: https://github.com/cynisca/weir-spec - [Weir — native mobile onboarding as code](/docs/): Weir is native mobile onboarding as code: a build-aware delivery and experimentation system for iOS and Android flows composed from app-owned components and authored through coding agents. - [Evidence status](/docs/evidence/): What Weir’s public local loop, qualified dogfood integrations, and remaining release gates do—and do not—prove today. - [You already have analytics. You still don't know where onboarding breaks.](/docs/why-observe/): Why Weir Observe exists for an app that already ships an onboarding flow and a general-purpose analytics SDK: a per-screen funnel in about fifteen minutes, numbers checked against ground truth, and nothing else switched on. - [Install the Weir skill / MCP server](/docs/install/): Register the Weir MCP server (or the identical CLI) in a coding-agent host, and verify the weir_* toolset loaded before you author a flow. - [Observe — a funnel for the onboarding you already shipped](/docs/observe/): Add Weir Observe to an app that keeps its own onboarding UI: one dependency, Weir.observe(...) with an ordered screens array, standard wire events, and the same funnel and trust reads as served flows. - [weir init — scaffold a new app in under a minute](/docs/init/): Run weir init once to write a valid weir/app.json, a starter specVersion-4 flow, and its built baseline bundle — a doctor-passing skeleton, fully local, no account. - [Quickstart for agents](/docs/quickstart/): Onboard an app with weir/app.json and weir_app_doctor, author a v4 flow spec, preview/walk/verify it, and ship it with weir_release. - [Author your first flow](/docs/first-flow/): A complete, validated specVersion-4 flow spec — variables, branching, interpolation — with the field-by-field walkthrough and the verify commands. - [Sandbox preview & branch verification](/docs/sandbox-preview/): Preview a flow in-browser with mock native personas, walk every branch headlessly, and read walk-report.json and the per-screen screenshots. - [Run everything locally](/docs/local/): The no-account path: use weir dashboard for a real local API and dashboard, then run the publish/manifest/OTA loop against localhost. - [Reference](/docs/reference/): Every MCP tool and CLI verb, every flow-spec node type, the frozen event vocabulary, and the current state of compliance linting. - [Asset tokens — see before you ship](/docs/assets/): Every moment.image/welcome.image token and every option.icon token, rendered, with an honest description — so you never ship one blind. - [Concepts](/docs/concepts/): The native component model, gating and the component manifest, the v4 delivery/signing model, the Firebase kill switch, experiments, and the instrumentation contract. - [Onboard an app](/docs/onboard-an-app/): Private-alpha v4 integration checklist for SwiftUI and React Native/Expo hosts: native baseline config, registry gating, signed updates, telemetry, and fallback proof. - [Monitoring](/docs/monitoring/): Use the dashboard and read tools to confirm a published flow is live, completing, and failing safely; publish/rollback and the experiment lifecycle run from the dashboard too. - [Experiments](/docs/experiments/): The experiment lifecycle: pre-register, serve real arms, observe traffic, then apply a fixed decision rule. - [Review a flow](/docs/review/): Preview a published flow's real bundle in a phone frame, test an arm through real bucketing, walk simulated purchases and permissions safely, and leave comments anchored to a screen or an element. - [Troubleshooting](/docs/troubleshooting/): Symptom → cause → fix entries for v4 authoring, native baseline packaging, registry/version gating, signed config fetch, kill switch, and local dashboard failures. - [Full site as one document](/docs/llms-full.txt)