Spec‑driven mobile test tool
A new open‑source agent, finalrun‑agent, was introduced as a spec‑driven mobile testing alternative that aims to handle flow failures beyond what Appium covers. (x.com) The project targets flaky UI flows and tries to make mobile QA more deterministic for CI pipelines. (x.com)
The final-run/finalrun-agent repository on GitHub describes an AI-driven CLI that runs repo-local YAML tests against Android and iOS targets and exposes the CLI as the npm package @finalrun/finalrun-agent; during installation it stages native driver assets under ~/.finalrun/assets/<version> and stores run artifacts under ~/.finalrun/workspaces/<workspace-hash>/artifacts. (github.com) FinalRun published a product blog post on March 17, 2026 outlining its roadmap and testing approach and announcing plans to make the tooling broadly available. (blogs.finalrun.app) A separate FinalRun benchmark repository reports a 76.7% success rate across 116 diverse real‑world Android tasks, which the project presents as a new state‑of‑the‑art result for agentic, vision‑based mobile test execution. (github.com) Community demos and writeups show the agent can analyze a repository to auto‑generate human‑readable tests, export them as.md files, and execute them via a CLI command such as./mobile-cli run./test/search.md; demo recordings and the agent-skills repo document that generation flow. (news.ycombinator.com) FinalRun’s public site and tooling notes list integrations with GitHub, Jenkins, and custom webhooks for triggering runs inside CI/CD pipelines, and the project positions itself to surface run artifacts and diagnostics back into developer workflows. (finalrun.app) Technical posts from FinalRun describe tackling specific flakiness sources such as random popups and outline a “see, think, act” loop the agent uses to check screen state and drive recovery actions during tests. (blogs.finalrun.app)
