Research prototype
Does your code match your requirements?
ConformGraph checks approved requirements against repository code, configuration and schemas. See what is supported, what is violated, and what still needs evidence.
Built for software teams reviewing implementation drift across a repository.
The problem we are working on
Requirements change.
Implementations drift.
A requirement can span a route, a handler, a configuration value and a database schema. Finding a relevant file is only the beginning: reviewers still need to establish whether those pieces satisfy the approved contract.
ConformGraph brings these checks into one auditable workflow. The intended users are engineering teams reviewing changes, maintaining services and investigating mismatches between project documentation and code.
- Implementation reviews: trace an approved requirement to concrete repository evidence.
- Change checks: identify configuration and structural contradictions before accepting a change.
- Evidence gaps: distinguish an unresolved question from a proven failure.
A small service. Three audit outcomes.
Inspect the evidence.
These are recorded outputs from the actual ConformGraph engine, run on an intentionally small synthetic fixture. Switching a scenario displays its saved result; it does not run a cloud model.
Requirements met
The route, effective configuration value and declared database column support all three obligations.
The recorded report could not be loaded. Use the JSON download to inspect it directly.
From documents to a reviewable result
Keep interpretation and verification separate.
AI can help interpret a requirement. Evidence must establish whether the implementation satisfies it.
- 1
Propose requirements
A provider adapter can turn project documentation into structured proposals.
- 2
Review and approve
A reviewer approves the requirement bundle before an authoritative audit.
- 3
Check the repository
ConformGraph extracts relevant facts and evaluates typed evidence.
- 4
Read the outcome
Inspect supported claims, concrete contradictions and unresolved evidence gaps.
Claude integration
Help with the reading.
Evidence for the decision.
The prototype includes an Anthropic Messages API adapter for proposing structured requirements. Proposed output is locally validated and requires review before use in an authoritative audit.
The published demo runs without an API call. The adapter is implemented; live Claude API validation is pending. API credits would support validating this integration and evaluating requirement proposals on representative project documents.
Claude does not issue ConformGraph's conformance verdicts.
About ConformGraph
Make implementation decisions easier to inspect.
ConformGraph is an early-stage developer tooling project founded by Lê Bảo Tấn Phong. Its purpose is to help software teams connect project requirements to reviewable implementation evidence.
The current stage is a research prototype. The local demonstration has been compiled and run on three controlled scenarios. It demonstrates route, configuration and SQL evidence checks, not universal verification of arbitrary code.
Dynamic or unsupported behavior can leave a requirement unresolved. An UNKNOWN outcome identifies that gap rather than claiming success.
Contact
Lê Bảo Tấn Phong
Founder, ConformGraph
For questions about the project, prototype feedback or potential collaboration, contact the founder.
Ho Chi Minh City
Public founder contact