Plan
Turn “build my app” or a pro feature brief into the next few scoped work units so the agent knows what to build and what to leave alone.
Docs
Public docs for guided verification while agents build.
AgentHelm helps a builder turn rough AI work into scoped build steps, attach to an existing agent-built project, keep the agent oriented while code is changing, and leave an evidence-backed handoff when work stops. Your code stays local. AgentHelm coordinates the build.
Guide
This is not a scanner waiting at the end. AgentHelm checks code while the agent works. No god files, no mystery diffs, no agent glob, no quiet drift, and no surprise auth or sensitive-surface sprawl.
Existing projects
Attach AgentHelm to an existing agent-built project. See what drifted. Fix it with your chosen local agent. Map god files, agent glob, tangled code, and risky auth, payment, signing, database, or MCP drift into focused fixes.
Database-aware checks
When database logic changes, AgentHelm can require evidence for migrations, query behavior, persistence assumptions, and data-bound features before handoff.
Handoff
When work stops, get a receipt that says what changed, what needs a human call, and where the next packet should start.
Docs on this static shell are product-facing. Private checking recipes, routing logic, signing material, and closed engine labels stay out of public pages.