Short version: AI can build order flow indicators. Our preliminary assessment puts a full documented equivalent of MZpack at 4-9 months and $5,000-25,000 in AI usage, plus €196-441 in recurring subscription costs. MZpack’s six-indicator package costs €399. The development estimate excludes your time and future maintenance and support; it is a planning estimate, not a measured rebuild.
A recent support conversation raised a reasonable question: if an AI coding assistant can build order flow indicators, what is the advantage of buying MZpack?
AI can help you write indicators. A small tool for a specific idea can be an excellent project. The buying decision becomes more interesting when the goal is to reproduce an entire product: all six indicators, every documented setting, the chart interactions, and consistent behavior across historical data, replay and live markets.
We do not claim that Claude, Codex or another capable coding agent is fundamentally unable to implement these functions. The advantage of buying MZpack is access to an existing integrated implementation, without first funding and completing that development project.
A working prototype is a useful first step
An assistant can implement a bid/ask ladder, accumulate delta, calculate a session volume profile and draw large-trade markers. Those results can already be useful. If that is the complete specification, a DIY indicator may be the right choice.
A full equivalent of MZpack Indicators has a much larger specification. Footprint includes imbalance and absorption settings, zones and statistics. Volume Profile adds TPO, developing and naked levels, stacked profiles, and interactive splitting and merging. Big Trade combines reconstruction, filters, iceberg detection and DOM-pressure analysis. Market Depth adds a historical heatmap, aggregation and liquidity migration. Volume Delta and Delta Divergence share calculations while adding their own presentation and signal rules.
The settings interact. A change to price aggregation can affect calculations, filters, markers and scaling. A recalculation must update the existing chart correctly. Saving a workspace and reopening it must restore the intended configuration. These behaviors need explicit implementation and verification.
The difficult part is knowing whether it is correct
A familiar-looking chart does not establish equivalent calculations. Two implementations can draw similar footprints while assigning trades differently, aggregating them into different reconstructed trades, or ending a support/resistance zone at different times.
Some features also depend on information that is available in one data mode and absent in another. A historical tick stream and a live quote stream are different inputs. An agent needs to understand those differences and preserve them in the implementation rather than invent missing information.
A meaningful comparison uses the same market recordings and settings for the original and the new implementation. Numerical values and signal events should be compared directly wherever possible. Screenshots help verify presentation and interaction, but cannot establish every internal state.
The engineering loop is straightforward to describe:
- Define the specification. Inventory features and settings, resolve ambiguities, and establish acceptance criteria.
- Capture reference results. Record calculations, signals, zone states and chart behavior from the original on reproducible inputs.
- Implement and compare. Let agents build features, investigate discrepancies and rerun regression checks after corrections.
- Validate the whole package. Test settings together, historical/replay/live behavior, workspace persistence, stability and performance.
Agents can perform much of this work. They still need a reliable NinjaTrader comparison environment and usable reference results. Giving them more time without a dependable test loop does not provide a dependable completion criterion.
We asked a coding agent to estimate the complete project
To put numbers behind the discussion, we prepared a development assessment with Codex, a GPT-6-based agent, using a read-only analysis of the MZpack Indicators source code and its direct dependencies, together with the public documentation.
The hypothetical developer would have the documentation and trial version, rather than our source code. The scope covers all six indicators and all documented settings. Rebuilding the strategy framework, licensing and code protection is excluded.
These are preliminary planning estimates with low confidence. They are not the results of an actual rebuild, an independent benchmark, or a measurement of Claude’s performance. The original trial was not tested during the assessment. The source-code analysis informed the scope; the time and token budgets remain estimates.
| Target | Estimated duration | Estimated AI usage |
|---|---|---|
| All documented functions and settings, with broad comparison against the original | 4-9 months | $5,000-25,000 |
| Exact behavior under every possible condition, including undocumented behavior | No defensible fixed deadline | No fixed upper bound without a spending cap |
The first target has a practical finish line: an agreed test matrix passes. The second asks for universal equivalence, which cannot be demonstrated with a finite test suite. Even a very close equivalent needs a defined scope and stopping rule.
The scenario includes an additional recurring subscription of €49/month, totaling €196-441 over 4-9 months. This is a budget assumption from the assessment, not a quoted Claude plan. USD and EUR are kept separate. It assumes existing hardware is sufficient and does not price the developer’s own time. Additional original-product access and other unpriced expenses may also be required.
The standard MZpack trial lasts 14 days. That is useful for evaluation and reference capture, but a multi-month comparison project needs a plan for continued access to the original.
When building your own makes sense
DIY development makes sense when you want a narrowly defined custom tool, enjoy the engineering work, need behavior that existing products do not offer, or value control over your own source code. AI makes those projects more accessible.
Buying an existing package makes sense when you want its full feature set available now and prefer to spend your time using the tools. MZpack provides the implemented calculations, settings and chart workflows together. You can evaluate that behavior in the trial instead of first creating your own implementation.
€399 compared with building a full equivalent
The MZpack Indicators package is €399 for all six indicators. In the assessed full-equivalent scenario, AI usage alone is estimated at $5,000-25,000, plus €196-441 for the assumed recurring subscription and 4-9 months of development and validation.
This comparison excludes the cost of maintaining and supporting the new product after development. It also excludes the monetary value of your own development time and any additional unpriced access or infrastructure expenses.
A small custom indicator can have a very different budget. For the complete documented package, however, €399 buys an existing implementation at a small fraction of the assessed development expense. AI makes building possible; buying MZpack lets you begin with the product already built.
Assessment basis: MZpack Indicators Agent Development Assessment, 12 October 2026, prepared by Codex, a GPT-6-based agent, from product source-code analysis and public documentation. References: MZpack documentation, trial terms, and OpenAI pricing used in the assessment. Actual development time and spending depend on scope, automation, model selection and test coverage.
