Before Command, Truth.
RepoTruth performs read-only, evidence-based repository diagnosis, the picture you consult before AI agents, migrations, modernisation or remediation begin.
It reveals architecture, dependencies, hidden coupling and structural drift as they actually exist, distinguishes what is directly observed from what is derived, inferred, ambiguous or still unknown, and keeps the evidence behind each observation. It interprets; your team decides. No production access, no credentials, and no change to your code.
Evidence visualization layer
Every Diagnostic Package is also a living projection. The Observatory renders the evidence, where dependency concentrates, where knowledge thins, where the unknowns cluster, and decides nothing. It shows; your people interpret.
Observatory ≠ Authority. A read-only visualization of evidence; it assigns no grades and decides nothing. Synthetic demonstration.
The shape of the problem
It happens the way every large system changes, quietly, one reasonable decision at a time.
Software begins simple. One clear idea, held in a few people’s heads.
Time passes. Deadlines are met. The system grows to meet them.
The architecture shifts to carry weight it was never designed for.
The people who understood the original decisions move on.
Documentation describes the intent, not what the code became.
Dependencies multiply until no diagram fits on one screen.
And gradually, no one fully understands the repository anymore.
Yet the team must still migrate it, modernise it, hand it to AI agents, or outsource its care.
If you have changed a repository like this before, without ever fully understanding it, you already know why RepoTruth exists.
RepoTruth restores understanding before change begins. It reconstructs what the repository actually is, marks what it cannot be sure of, and gives your people the evidence, before anyone commits to a change.
The problem
Before the migration, the modernization, the acquisition, the picture is rarely complete.
By the time an organization commits to changing software, the decision is usually irreversible, time-boxed, and based on partial knowledge. The people who understood the system have often moved on. The documentation describes intent, not reality. And the cost of investigating the wrong thing first is measured in quarters, not days.
These aren’t engineering problems. They’re decision problems, and they arrive before any code changes.
The category
Not code generation. Not a dashboard. Not an autonomous repair agent. Evidence-governed visibility, before action.
RepoTruth is evidence-governed repository diagnosis: read-only observation of a repository as it actually exists, with the evidence and the uncertainty kept visible. Software Decision Intelligence remains a useful way to describe what the diagnosis is for, but the product itself is the diagnosis.
Existing tools answer operational questions: is it running, is it secure, is the build green. None answer the question that comes before a change: given what we can actually observe, what is really in here, how confident is each reading, and what don’t we know yet? That question has never had a home. RepoTruth gives it one.
RepoTruth is built for the decision, and only for the decision. It produces evidence, interpretation, and priorities, and leaves the conclusion to you.
In one glance
The capability and the boundary, side by side, so there is nothing left to assume.
The deliverable
One complete, decision-ready picture, built to be read by the people who decide.
Every engagement produces a Diagnostic Package: a structured, evidence-based account of the system, what deserves investigation, how confident we are, and what remains unknown. It is not a score. It is not a verdict. It is the picture an executive and an architect can sit down with, together, and act on.
Why the package exists A decision deserves more than a dashboard and less than a guess. The package gives your people everything they need to decide, and stays honest about the limits of what it knows.
Explore the Diagnostic PackageObserved · deserves investigation · remains unknown, readable in one sitting.
The system’s actual structure and strata, observed rather than assumed.
Every observation, cited, each carrying its own confidence.
Where to investigate first, ordered by investigation urgency, never by “severity.”
What we could not determine, surfaced with equal prominence and never hidden.
How firmly each finding is held, and why.
The evidence, made visible.
Exactly what was analyzed, what was not, what remains unknown, and the boundaries we work within.
The exchange
A bounded snapshot in. A complete, decision-ready picture out.
Everything remains read-only. Nothing is executed. Nothing is modified.
Who it's for
For the people who carry a major software decision, and the consequences of getting it wrong.
Inherited complexity is hard to see before you commit to changing it.
RepoTruth reconstructs the real system from a snapshot, before modernization begins.
Outcome: enter the program knowing where the risk and the unknowns actually sit.
Documentation describes intent; the running system has drifted from it.
RepoTruth reveals hidden dependencies, authority concentration, and unstable boundaries.
Outcome: an observed baseline to redraw the architecture from.
AI systems inherit the environment they are placed into, including its unknowns.
RepoTruth shows the structure and the risk surface that automated change will meet.
Outcome: know what to investigate before AI touches the codebase.
You must form a technical view without touching the target's environment.
RepoTruth surfaces structural unknowns from a read-only snapshot, independently.
Outcome: a defensible evidence base before a strategic commitment.
Refactoring and decomposition begin blind when complexity is invisible.
RepoTruth maps where inherited complexity concentrates, before you cut.
Outcome: decompose from a map, not from assumptions.
When to use it
RepoTruth is built for the moments where the cost of being wrong is highest.
Understand where automated change carries the most risk, so human review precedes it.
See the real structure, and its unknowns, before committing a multi-year program.
Identify where change is likely to propagate, and where to look before you move.
Establish the observed baseline before you redraw it.
Build a defensible, read-only evidence base for the diligence file.
Form an independent technical view of the target, without touching the asset.
Assess delivered software objectively, on evidence rather than assertion.
Decision intelligence
It never answers them for you. It surfaces the evidence so your people can.
RepoTruth surfaces evidence. Humans make decisions.
Outcomes
Used before the largest, least-reversible software decisions, for what it gives the people who make them.
The engagement
Seven steps. No production access at any of them.
You provide a bounded, read-only, point-in-time copy. No live connection, no credentials.
The snapshot is received into an isolated boundary; any sensitive material is flagged for your attention, never used.
The system is reconstructed as it actually is: structure, dependencies, history, and the limits of what can be seen.
Evidence, signals, investigation priorities, confidence, and preserved unknowns are derived from the reconstruction.
Everything is assembled into one structured, decision-ready deliverable, with its own statement of limits.
An explorable projection of the reconstructed reality is delivered privately, drillable to the evidence beneath.
Your people read, interpret, and decide. RepoTruth brings the evidence and the priorities; the conclusion is theirs.
No code execution. No system access. No automation.
Constitutional guarantees
Six permanent guarantees, not settings, not promises. Properties of how RepoTruth is built.
Most platforms ask for access and ask you to trust them with it. RepoTruth asks for none. Its boundaries aren’t policies that can be reconfigured, they’re the architecture itself. Here is what that means, permanently.
We write to nothing in your environment. There is no path by which RepoTruth can change your code.
We analyze an inert, point-in-time copy, never a live or running system.
We never connect to production. Not to observe, not for any reason.
We store, request, and reuse no credentials, because we connect to nothing that needs them.
We execute, deploy, and remediate nothing. RepoTruth cannot act on your systems.
Every output is built for a human to review and decide on. RepoTruth reaches no conclusion of its own.
The unknown
An unknown isn’t a gap to hide. It’s one of the most useful things we can give you.
Most tools quietly fill in what they can’t determine, or omit it entirely. RepoTruth does the opposite. Where the evidence runs out, we record an Unknown, explicitly, prominently, as a first-class part of the picture. Ownership that can’t be established. Behavior a snapshot can’t reveal. Rationale that was never written down. Surfaced, never assumed away.
An unknown is not low risk. It is un-assessed. A region full of unknowns isn’t safe, it’s precisely where your people should look. The most valuable page in a Diagnostic Package is often the one that lists, exactly, the questions worth asking next.
We would rather tell you the truth about what we can’t see than hand you a false sense of completeness.
The interface to software reality
It is an explorable projection of repository reality, reconstructed from a bounded snapshot.
The Observatory reveals where dependency concentrates, where change churns, where ownership is uncertain, where knowledge thins, and where the unknowns cluster, the regions that most deserve a human’s closest reading.
It shows; it assigns no grades, computes no score, and decides nothing. You interpret, and you decide.
Proof
We ask you to trust the method, not the marketing.
RepoTruth operates against a written constitutional standard that governs what it may produce and what it may never claim. Your security and legal teams can read it.
See a complete Diagnostic Package and Observatory assemble on a synthetic codebase, before you share anything of your own. No real repository required.
Every package states what it analyzed, what it did not, and what remains unknown. The limits are part of the deliverable, not the fine print.
RepoTruth is designed to be reviewed, challenged, and decided upon by your people. Its outputs cite their evidence and carry their confidence, so they can be checked.
We don’t publish fabricated customer stories or invented metrics. The proof is the method, the manifest, and the package you can hold.
Founding Expert Review
A small, invitation-only program for experts willing to pressure-test the discipline.
Questions
Every answer reflects what is true today. Planned capabilities are not presented as operational.
No. RepoTruth is read-only by architecture. There is no path by which it can write to, execute, or change your code.
No. It works from a bounded, inert, point-in-time snapshot. It connects to no live or running system and stores no credentials.
Yes. RepoTruth is designed to run in your own environment, and its analysis path is offline-capable, your source can stay under your control.
No. It observes the snapshot without executing it. Diagnosis is not execution.
Uncertainty is preserved as a first-class finding. Observations are marked directly observed, derived, inferred, ambiguous, or unknown, and what cannot be determined is surfaced, never assumed away.
No. The Observatory is a navigable projection of the diagnosis, not its source of truth and not terminal access. The evidence is the authority.
No. RepoTruth brings evidence, priorities and preserved unknowns to a human review. Every conclusion is reached by your people.
Ordinary tools maximise findings and often separate a finding from its evidence. RepoTruth prioritises supportable observations, keeps evidence and provenance attached, preserves ambiguity and unknowns, and stays read-only.
No. RepoTruth does not remediate, deploy, or act on your systems. It diagnoses; your team decides what to do.
Repository content is used only to produce your diagnosis. It is not used to train models. Where a hosted evaluation is used, retention is controlled and stated in the engagement.
A read-only evaluation, scoped to the decision in front of you.
Tell us the decision you’re facing. We’ll scope a read-only Repository Evaluation to it, snapshot-based, with no production access and no obligation to act on what we find. Your people will know what is actually in there, how confident each reading is, and what is still unknown.
No credentials. No production access. We never change your code.