Capawesome lets AI agents trigger app builds, releases, and rollbacks
Mobile app build-and-release platform Capawesome now lets AI agents perform tasks that previously required a human in its dashboard: triggering native iOS/Android builds, signing them, shipping to TestFlight or Google Play, pushing live updates, and rolling back a bad release, all from a prompt. Every dashboard action is exposed as a CLI command with , so any agent or script can parse results reliably. Eight official skill packs, written in plain Markdown, install with one command (`npx skills add capawesome-team/skills`) and work with Claude Code, Cursor, Codex, Windsurf, or a custom runtime.
Non-interactive auth tokens support headless or CI environments. The security design is the centerpiece: agents never touch raw credentials. s, keystores, and store API keys sit in an encrypted vault; the agent only references them by name, and the actual keys are unwrapped solely inside isolated build environments, never appearing in the agent's memory, prompts, or logs.
Tokens can be scoped with control (RBAC), so an agent can be allowed to ship to staging while production deployment stays a human decision. Because builds run in the cloud, even a Linux-based agent can trigger an iOS build without owning a Mac.
Key points
- Every Capawesome dashboard action (build, sign, deploy, rollback) can now be triggered by an AI agent via prompt
- Eight official skill packs install with one command and work across Claude Code, Cursor, Codex, and Windsurf
- s and API keys stay in an encrypted vault; agents reference them by name and never see raw values
- RBAC lets teams scope agent tokens so staging deploys are automatic but production stays a human decision
- Cloud-based builds mean a Linux agent can trigger an iOS build without needing a Mac