An agentic dev stack for Claude Code

The dev stack that ships itself.

Autonomous builds with maker/checker review, hardened guardrail hooks, deterministic memory, and the manual that explains why every rule exists - packaged from a working senior setup, not a prompt pack.

Get the kit - $49 Hire the pipeline - $1k/mo

faur-kit - ninety seconds to a working forge

$ unzip faur-kit.zip && cd faur-kit && ./install.sh ✓ settings.json merged - 0 conflicts, backup kept ✓ plugins pinned: faur · insights · ponytail $ ./doctor faur-kit doctor [ok] jq installed [ok] claude CLI installed [ok] settings.json valid JSON [ok] kit hooks present [ok] kit hooks executable [ok] kit hooks registered doctor: 6 ok, 0 failing

Installer merges - never overwrites - your existing settings. Rollback is one command.

01 - Receipts

2 home-built plugins + curated stack 8 guardrail hooks + doctor 15/15 retrieval bench (dev corpus) maker/checker on every build 0 deps - stdlib + your CLI

every number on this page is re-derived from the shipped files at package time — tools/claims_audit.sh

02 - It built this

Watch the forge work.

Terminal recording: unzip faur-kit, install.sh merging settings, doctor going green — the install cut - real terminal, real time

Terminal recording: the faur pipeline building a tested todo CLI in parallel batches — the pipeline building a tested todo CLI

the same pipeline built, deployed, and sells this page.

03 - Bill of materials

What's in the crate.

Click a part for its receipt - a real command and what it prints.

Part	What it does	Condition
	Autonomous build engine - chunks a backlog, builds in parallel batches, every batch adversarially verified before it lands.	daily use
> /faur-auto shop-app "ship the next two admin screens" batch 1/3 · 4 agents · maker/checker on
	Severity-laddered review after every task; security lens auto-triggers on sensitive paths.	battle-tested
$ head -8 manual/02-review-pipeline.md Single reviewer after every implementation task. Panel review (2 reviewers, add a 3rd if a security path fired) triggers on ANY 2 of: …
	Deterministic fact memory - exact retrieval without burning context. Ships empty; your facts stay yours.	bench 15/15
$ ./second-brain/brain.py find "how do I roll back the install" ## 040f61 (score 4) - rollback is one command Evidence above is deterministic (zero model calls in retrieval).
	Per-project usage reports: where your tokens and hours actually go.	daily use
> /insights # report lands in ~/.claude/insights-reports/<project>/report-<date>.html
	Guardrails - broken-config detection, git safety rails, session health. doctor proves the install.	hardened
$ ls hooks/*.sh check-status-line.sh check-uncommitted.sh check-visual-fidelity.sh gh-mutation-block.sh scan-subagent-prompt.sh session-health.sh stop-orchestrator.sh write-md-warning.sh
	What the guardrails actually did for you this week - blocked mutations, nudges, per-hook counts. The kit showing its receipts.	fixture-tested
$ ./receipts # sample window - the fixture shipped in tests/ gh-mutation-block 2 blocked last: gh api -X POST /repos scan-subagent-prompt 1 blocked last: git mutation in Task prompt write-md-warning 1 nudge last: prefer Edit over Write for large file README.md total: 4 interventions across 3 guardrails
	The CLAUDE.md architecture with the why behind every rule - templates you adapt, not magic you inherit.	documented
$ ls manual/ 00-why.md 01-global-claude-md.md 02-review-pipeline.md 03-memory-routing.md 04-model-routing.md

04 - Pricing

Buy the forge, or hire the smith.

The kit $49 one-time · launch price

in your first 15 minutes: installer merges (never overwrites) → doctor proves it → guardrails live

Full stack + installer + the manual
Updates ship while the forge runs - this field re-tools weekly, so does the kit
14-day refund if doctor won't go green on your machine

Get the kit

The pipeline · 2 slots $1,000 / month · cancel anytime

Your backlog, shipped weekly by this stack - human-reviewed before merge
Kit included; so is the person who built it
Slot holds priority on new capabilities as they land

Book a slot

05 - Straight answers

Won't this be obsolete in six months?: Parts of it, probably - this field re-tools weekly. That's why it costs $49 and not $499, why updates ship from live daily use, and why the retainer exists: you're buying the current frontier, not a forever tool.
Will it actually run on my machine?: Every release passes a fresh-machine install test before it ships. doctor verifies your install end-to-end - if it won't go green, you get your money back.
Does it phone home or read my code?: No. Everything runs locally against your own Claude Code account. The memory system ships empty. The only network calls are the ones your CLI already makes.
What support do I get for $49?: The manual, the README, and the issues tab - honestly. Humans come with the pipeline slot.