NEWv0.1 · LangGraph + Claude · Read the architecture

From goal to production
in 30 minutes.

A deterministic LangGraph pipeline deploys 20+ specialist Claude agents — CTO, architect, security reviewer, domain engineers — to ship production fintech, crypto, and AI products.

LangGraphClaude OpusClaude SonnetClaude HaikuHardhatFoundryPythonTypeScript
minctrl · workflows · thread-3293eb0c
LIVE
Pipeline — fintech · neobroker
0.0s
01
InitializeRouter · Config Loader
02
CTO Debate
03
Arch Design
04
Security Gate
GATE
05
Domain Team
06
DevOps
07
QA
08
Ship
streaming consoleagents · 1
[00:00] ▸ detected: vertical=fintech product=neobroker
Stage
Initialize
Tokens
0
Cost
$0.00
Agents
16
53
Pipelines shipped
20+
Specialist agents
18
Product types
$1.47
Avg cost per build
~30m
Median time-to-ship
8
Fixed stages
Why it's different

Not a code generator.
An automated engineering team.

Deterministic pipeline, not a dice roll

LangGraph state machine — 8 fixed stages with checkpoints, security gates, and a human-readable audit trail. Every run follows the same proven process.

Domain expertise, not a generic model

Separate agent teams per vertical. Fintech agents know FIX, MiFID, pre-trade risk. Crypto agents know reentrancy, oracles, MEV. Not templates — specialists.

Full automation, not just code

The board populates itself. The wiki writes itself. Security reviews on its own. CI/CD scaffolds itself. Code is one output of seven.

Under the hood

Built on LangGraph.
Deterministic. Auditable.

A LangGraph state machine coordinates every agent through 8 fixed stages with built-in checkpoints. No black box — full agent log, cost breakdown and audit trail for every run.

Stage 1 · Initialize

Detect pipeline type, load domain config, bootstrap state

RouterConfig Loader
Inputs
→ Natural language goal
Outputs
← Detected pipeline type
← Domain config
← Initial state
Model
Haiku
Avg cost
$0.01
per stage
Strategic agents
Claude Opus

Deep reasoning: architecture debate, security review, meta-critique

CTOSystem ArchitectSecurityQA LeadMeta Critic
Domain agents
Claude Sonnet

Specialists per vertical — fintech, crypto, AI, infrastructure

Domain SpecialistProduct OwnerAPI DesignerDB EngineerTech Lead
Execution agents
Claude Haiku

High-frequency tasks: DevOps config, cost tracking, formatting

DevOpsPrompt EngineerReliabilityCost Optimizer
Three deep verticals

Fintech. Crypto. AI.

Not a general-purpose tool — each vertical has dedicated domain agents carrying production-grade knowledge. FIX protocol. Reentrancy guards. Hybrid retrieval.

src/risk/RiskCheck.ts● generated by Fintech domain agent
// RiskCheck.ts — generated by fintech domain agent
export async function checkBuyingPower(
account: Account, qty: number, price: number
): Promise<RiskResult> {
const required = qty * price +
estimatedFees(qty);
if (account.buyingPower < required) {
return { ok: false, reason: "INSUFFICIENT_BP" };
}
// MiFID II pre-trade log
await mifidLog({ account, qty, price });
return { ok: true };
}
Human-in-the-loop

Pipeline pauses.
You review. Then it ships.

Choose your level of control. Gates pause the pipeline at critical decision points — a kanban card appears with a GitHub link so you can review the code before approving.

Two checkpoints: architecture sign-off and final ship approval.

Start
CTO
ArchitecturePAUSE
SecurityMANDATORY
Domain Team
DevOps
QA
ShipPAUSE
Done
auto-executingpause for reviewmandatory gate
Full transparency

Watch every agent think.
Every decision logged.

The execution graph shows every agent, every stage, every output in real time. Streaming console gives a live feed of agent reasoning. Full cost breakdown per agent, per run.

Real-time agent execution graph
Streaming console — watch agents reason as they work
Token-level cost breakdown: per agent, per stage, per run
Download all output artifacts: code, tests, docs, CI/CD
Agent execution graph
Platform

Everything in one workspace

Six integrated modules, built around the agent pipeline. Each one kept in sync with the live run state.

01

Dashboard

Live workflow status · recent runs · quick switcher

02

Projects

Pipeline type · goal & config · run history per project

03

Workflows

Execution graph · streaming console · artifacts

04

Board

Auto-populated by domain agents · status tracking

05

Wiki

PRD · arch · API docs · data models · always synced

06

Analytics

Per-agent cost · token usage · traditional-team compare

Benchmark

Minctrl vs. a traditional team

Metric
Minctrl
Traditional team
Time to ship
~30 minutes
2 – 6 weeks
Cost
$0.50 – $2
$5k – $20k
Output
PRD · Arch · Code · Tests · Wiki · CI/CD
Code only
Review trail
Every decision logged
Slack history
Security
Mandatory gate on fintech/Web3
Best effort
3 pipelines running now

One goal.
20 agents. 30 minutes.

Ship your next fintech, crypto, or AI product with a full engineering team's worth of output — for less than a cup of coffee.

✓ No credit card✓ Full audit trail✓ GitHub-ready output