Entropix gives AI coding agents a process, the know-how to follow it, and a reviewer that holds the work to what was asked.
Paste notes or attach a file — never a form. The PRD's agent names it, drafts it in a fixed shape, and keeps every change as a new version. Ask for a change in the PRD's own chat while the first draft is still being written; your message waits for it.
The Entropix plugin for Claude Code turns a PRD number into a working session: it fetches the PRD, asks your agents what they know, clones what it needs and opens a merge request that names the PRD. The portal only reads.
work-a-runconsult-agentsbuild-agentsuggest_ruleEvery merge request that names a PRD is reviewed against its acceptance criteria. A finding is not a failure — it is a change requested, explained in the review's own words. Pick an angle to see what it checks.
Does the change meet each acceptance criterion in the PRD? Every criterion is marked met or not, with the evidence.
Will the shape of the change hold as data, traffic and the team grow?
The conventions your agents’ know-how records for this codebase — naming, errors, tests, structure.
Inputs, secrets, permissions and the boundaries between organisations and users.
Is there a simpler or existing path the change walked past?
Is it harder to read, test or change than it needs to be?
Callers, data and behaviour other people depend on.
Import an agent from Git, or build one from a Claude Code session with build-agent.
A docked chat on every agent, with a context bar and automatic compaction for long threads.
Mention the agent in a thread to ask, or write goal: to hand it work.
Point an agent at your clusters. It reads freely; every change asks first.