Closed source · In development · Private testing

The software company that lives in your repo

LARPCO is a closed-source autonomous agent-company framework currently in development and private testing. It spawns a full cast of named AI agents who plan work, debate approaches, and seek approvals before execution.

Structured autonomous workflows, currently in controlled testing.

Meet the team See how it works

Dottie → Orchestrator

Xanthe → PM Agent

Skye → Design Agent

Blaze → Eng Lead

Alma → Loop Reviewer

Bindi → Engineer Worker

Velma → QA Reviewer

Pixie → Feature Scout

Dot → Learning Agent

Dottie → Orchestrator

Xanthe → PM Agent

Skye → Design Agent

Blaze → Eng Lead

Alma → Loop Reviewer

Bindi → Engineer Worker

Velma → QA Reviewer

Pixie → Feature Scout

Dot → Learning Agent

What it is

Structured automation.
Real engineering.

LARPCO spawns a full cast of AI agents inside your GitHub repo. They have roles, titles, and a chain of command. Under the costume, the architecture is solid. The bit just makes it more fun to run.

🏢

A company in your repo

Orchestrators, PMs, engineers, reviewers, scouts. They plan sprints, write PRDs, debate approaches, open PRs, and post standups to Slack. You're the CEO. That's your entire job.

🔁

Ralph Loops

Every engineer agent runs in a fresh context window with exactly one task. The agent that writes code never performs final review on the same change. Separation of concerns is enforced by design.

🧠

Compounding memory

Every completed loop writes its learnings back to a shared ruleset. Bugs get documented so they can't recur. The system gets smarter through each internal testing cycle. Dot remembers everything.

💬

Slack-first human loop

Approval checkpoints via Slack. You approve scope, not implementation. A reaction emoji kicks off the pipeline. You get a ping when something needs a human. That's it.

🚦

Pre-execution plan review

Alma catches bad plans before any engineer spends tokens on them. A bad loop plan wastes every token downstream. Plans are checked before workers run. Always.

⚡

GitHub Actions runtime

No proprietary platform. No vendor lock-in. The pipeline runs on standard CI infrastructure in a private testing setup. One script. Tight feedback loops. Still evolving.

Company snapshot

Quick facts for reviewers

A clear view of who we are, what stage we are in, and how teams can engage right now.

Company

Punkyin · AI-first product company (not an agency or consultancy)

Founded

2026 · Gold Coast, Australia

Product stage

LARPCO is in active development and private testing

Availability

Closed source today · Early access intake open

Cloud + AI

Built and tested on Google Cloud, with planned expansion on Google AI models

Contact

b@punkyin.com · LinkedIn profile public

The cast

Agent roles and responsibilities

The current LARPCO build uses internal codenames for each role. Each agent has a clear scope, explicit handoffs, and a defined review boundary.

Name	Role	Responsibility
Dottie	Orchestrator	Coordinates pipeline state, routes tasks, and enforces execution order.
Xanthe	PM Agent	Converts goals into scoped requirements and acceptance criteria.
Skye	Design Agent	Produces implementation-oriented design and interface guidance.
Blaze	Eng Lead	Breaks requirements into atomic implementation tasks.
Alma	Loop Reviewer	Validates plans before execution to reduce downstream rework.
Bindi	Engineer Worker	Implements scoped code changes in isolated task contexts.
Velma	QA Reviewer	Performs independent QA review and requests fixes when needed.
Pixie	Feature Scout	Proposes next-step opportunities based on recent outcomes.
Dot	Learning Agent	Captures learnings and updates persistent operating guidance.

How it works

The full pipeline

A task enters with human approval and progresses through planning, implementation, and independent review. You approve scope; the system executes the workflow.

Trigger

You approve a task on Slack

Human approval triggers routing and execution of the next workflow stage.

↓

Upstream

Xanthe — PM

Translates a goal into a PRD with acceptance criteria. Scoped to prose. No code.

Upstream

Skye — Design

Produces implementation-focused design guidance from the scoped requirements.

↓

Planning
Blaze — Eng Lead
Reads requirements and design guidance, then produces atomic task briefs for workers.

Gate
Alma — Loop Reviewer
Independent reviewer validates plans for ambiguity, missing criteria, and architectural risk.

↓

Implementation

Bindi — Workers

Runs in fresh context, implements scoped changes, and includes tests where required.

Review Gate

Velma — QA

Independent QA pass verifies correctness before progression.

↓

Output

Internal test run · Passing checks · Standup on Slack

Execution summary is posted, follow-on opportunities are proposed, and learnings are recorded.

The orchestrator does not implement code. It reads workflow state, routes tasks to the correct role, and keeps decision context explicit. Core orchestration lives in run-pipeline.mjs. The system currently runs in private testing while architecture and safety checks are refined.

Positioning

The honest comparison

GSD says "no enterprise theatre." LARPCO says "we exclusively do enterprise theatre." The PRs still merge.

	GSD	LARPCO
Vibe	Minimalist, efficient	Role-based workflow with explicit handoffs
Structure	Commands you run	Company that runs itself
Human role	Developer	CEO
Agents	Subagents, anonymous	Named cast, roles, chain of command
State	.planning/ directory	ORCHESTRATOR_STATE.md + git
Approvals	Interactive prompts	Slack reactions
Tone	"No bs"	Process clarity with strong guardrails
Works?	Yes	Also yes

The bit is real

Under the costume

Every decision in the system was made to prevent a specific failure mode. The comedy is in the framing, not the output.

Fresh context, every task

Agents accumulate bias and drift when they carry context across unrelated tasks. Every worker spawns with a clean window and exactly the context it needs.

Implementer ≠ reviewer

An agent cannot objectively review its own output. Velma and Alma are always separate instances from the engineer. Non-negotiable.

Specialization over generalism

Xanthe writes better PRDs when she's never touched code. Blaze plans better when he doesn't also have to implement. Specialization compounds.

Plans gate execution

A bad loop plan wastes every token downstream. Alma exists to catch under-specified plans before any engineer spends tokens on them.

Learnings write back to memory

Dot updates the shared ruleset after every completed loop. Bugs discovered during testing get documented so they cannot recur. The system improves with every task.

Humans approve scope, not implementation

You decide what to build and approve task scope. The agents decide how to build it. This keeps you strategic and the pipeline fast.

Early stage

We're early —
and that's the point

LARPCO is in active development, currently in private testing, and closed source for now while we harden the core system. We're early, learning fast, and opening conversations with teams interested in shaping early access.

b@punkyin.com →

Join early access

Requests are reviewed weekly. Prefer email instead? Write to b@punkyin.com.

Who built this

The human in the loop

Built by a real company with a real founder, a clear model, and an early product currently in testing.

Punkyin

Punkyin is an AI-first product company founded in 2026 on the Gold Coast, Australia. It is bootstrapped, solo female-founded, and focused on building tools for developers.

Punkyin is not an agency or consultancy. The long-term model is a source-available core with a paid hosted offering for teams that want managed infrastructure.

The current system is being built and tested on Google Cloud, with plans to expand use of Google AI models.

Flagship product: LARPCO.

Brianna Malcolmson

Founder

Brianna is a cybersecurity product leader who has worked across cloud security, incident response, red teaming, and product management. She founded Atlassian Beacon and is an inventor on U.S. Patent 11,895,130 for proactive suspicious activity monitoring. She was also a full-ride golf scholarship athlete at Penn State.

LinkedIn b@punkyin.com

FAQ

Current status, plainly

Short answers to the questions reviewers and early partners ask first.

Is LARPCO open source?

Not yet. LARPCO is currently closed source while we harden architecture, safety checks, and operating workflows.

Is LARPCO in production?

No. It is in active development and private testing. We are not presenting it as generally available production software today.

Who is building it?

Punkyin, a bootstrapped solo female-founded company led by Brianna Malcolmson.

How can teams engage?

Join the early access list or email b@punkyin.com. Requests are reviewed weekly.