Read-only analysis
Source is parsed as data. Imported code, scripts and dependencies are never executed.
Bring requirements, code, tests and CI into one workspace. Inspect the evidence behind delivery claims—and make the next release decision with clarity.
Loading the captured source…Captured revisionSource stays inspectableOne place to connect what you planned, what the repository contains, and what still needs review.
Bring requirements and acceptance criteria into the same view as the code. See the candidate implementation and tests behind each one.
The product preview uses captured source records. You can also open the working workspace.
Open workspaceStart with source evidence. Make the reasoning inspectable. Keep the final acceptance decision with a person.
Inspect the Azure model review record below. Live Copilot defaults to evidence retrieval; model explanation is an optional, separate setting.
Reading the saved review status and any recorded response. This does not start a model request.
This recorded review is separate from the live workspace’s optional model setting. No completed result is assumed from configuration.
Read the supported files from one immutable revision.
Extract symbols, imports and named test declarations.
Match criteria with candidate source and test records.
The assistant answers from the saved analysis records.
Implementation, matching tests, confirmed criteria and passing CI for the same commit are all required for “verified.” The explanation mode cannot approve a requirement.
Deterministic retrieval uses the analysis already on record. It can locate requirement evidence, explain missing checks and return citations without calling a language model.
This switch explains the options; it does not enable a provider or send source.
Coverage measures the presence of evidence. It does not estimate the probability that software is correct.
For the people who need to explain what the software does—and what still needs checking.
Different review perspectives. One evidence workflow.
Bring the delivery conversation back to inspectable evidence.
A green CI result is one part of the decision. You also need to see the requirements, their supporting evidence, and the review still open.
Read the recorded CI result alongside the reasons that keep the project under review.
Follow a requirement to its candidate implementation and named test declarations.
The Nano ID checklist remains unreviewed. Passing recorded CI does not approve it.
A release summary with evidence and unresolved review reasons.
Your task tracker records the plan. CI records a run. SofiCore connects requirements to the captured evidence and review state. This MVP supports one owner and one repository; these are intended uses, not separate permission roles.
Explore the supported scopeWe applied SofiCore to a real public JavaScript repository. These findings come from one captured revision, with every review decision still visible.
Implementation, criteria mappings, named tests and recorded CI contribute to this score.
This is a traceable review result. It is not a claim of defects in Nano ID or approval from its maintainers.
Reading the saved requirement findings…
Select a documented behavior, inspect the source and tests, then verify the saved files. You’re using the same captured record as the full workspace.
Preparing source links, requirement matches and the file receipt.
Independent public-source analysis. Criteria are derived from Nano ID’s README and remain unreviewed; the maintainers have not endorsed this study.
Read the full case studyThe trail becomes available when the captured case finishes loading.
Explore the public workspace first. The MVP is designed for one owner reviewing one small JavaScript or TypeScript repository.
Source is parsed as data. Imported code, scripts and dependencies are never executed.
Saved workspace access is checked on the server. The owner controls active source records.
Default Copilot uses project records. Optional external model requests require configuration and consent.
The repository adapter and owner workspace are implemented. Hosted account and deployed import acceptance are still pending; private access needs a registered GitHub App. AWS infrastructure is documented but not provisioned. The public case is available in this evaluation build.
View current configurationClear definitions make the evidence useful.
Explore the casesCoverage measures the evidence inventory: candidate implementation, mapped criteria, test declarations, and recorded CI. Verification also requires explicit acceptance review. In the Nano ID snapshot, all seven requirements have evidence, and all seven still await that review.
No. This is SofiCore’s independent analysis of public open-source files at one immutable commit. SofiCore derived the checklist from the upstream README. The maintainers have not supplied or approved the checklist, findings, or release decision.
No. The analyzer found 55 active named test declarations in source. It does not execute the imported repository. Four successful GitHub Actions runs were separately recorded for the captured commit; they are linked in the case study.
No. Reanalysis reads the saved snapshot and updates the analysis time. The source capture date and commit stay fixed. Observing a newer upstream revision requires a new import or capture; this MVP does not provide continuous synchronization.
It compares the saved file contents with the SHA-256 hashes recorded in the receipt. A match proves consistency with that receipt. It does not establish human approval, upstream authenticity, CI execution, or production readiness.
Default evidence mode does not. An operator can configure an optional model adapter; using it requires explicit opt-in and sends a bounded evidence packet. The public demo stays in evidence mode. Read the security and privacy pages for retention and provider boundaries.
The repository adapter and owner workspace are implemented, with setup documented in the source package. This preview has not completed hosted account and deployed repository-import acceptance. Private access also needs a registered GitHub App. Start with the public case and check Operational status before planning a deployment.
Atlas Commerce is a separate, runnable training application with deliberately incomplete requirements. It helps demonstrate gaps and blocked review states. It is not a customer project, and its remote CI is unknown.
Explore a real project. Follow a requirement. See what still needs review.
No personal repository needed to explore the case