Player profile
Sandcastle
A TypeScript library and CLI for orchestrating AI coding agents in isolated sandbox environments with branch strategies, hooks, logs, templates, and merge-back workflows.
Profile
Where it fits
Sandcastle is an MIT-licensed TypeScript library and CLI for orchestrating AI coding agents in isolated sandboxes. The README frames the core loop as: invoke agents with `sandcastle.run()`, sandbox the agent with a configurable branch strategy, then merge the commits made on the branch back into the host repo.
It belongs in Open Orchestrators because it provides a programmable control layer around coding agents: sandbox providers, worktrees, branch strategies, host and sandbox hooks, prompt templates, dynamic context commands, logging, review-pipeline templates, and support for providers such as Claude Code, Codex, and OpenCode.
Builder web analytics
Measure projects built with Sandcastle
Sandcastle can run coding agents in isolated sandboxes and merge their commits back into a project. Agent Analytics measures whether the shipped change improved the user-facing surface after that agent work lands.
Instrument the deployed surface affected by Sandcastle-managed commits. Agent Analytics reads product and web events after the branch ships; it does not replace Sandcastle logs, sandbox traces, branch state, or provider output.
First loop to measure
- a Sandcastle run asks Claude Code, Codex, OpenCode, or another coding agent to change a website, docs flow, onboarding path, app surface, demo, or experiment
- the changed surface reports visits, sources, CTA clicks, signup, activation, retention, or task-completion events to Agent Analytics
- a follow-up Sandcastle run or developer agent fetches Agent Analytics results after deployment
- the next sandboxed run is scoped from measured user outcomes instead of only code-level completion
Copyable prompt
Use Agent Analytics for this project. If event reporting is missing, add the tracker and report events for this project surface, including Sandcastle-managed page, docs path, traffic source, CTA click, signup, activation event, retention signal, or shipped experiment. Verify events are arriving. Then fetch the last 7 days and compare them with the prior 7 days. Tell me which Sandcastle-managed page, docs path, traffic source, CTA click, signup, activation event, retention signal, or shipped experiment moved users toward value, where users dropped off, which sources mattered, and what my agent workflow should improve next.