Upload a ZIP of legacy Java, JSP, and SQL — no running system needed. In about thirty minutes, Codolith hands back a queryable knowledge graph, reviewable blueprints, four audit-grade PDF reports, and a chat that cites its sources down to file:line.
How it works
Upload a ZIP of the source. No build, no runtime, no access to your servers — Codolith reads the code as-is.
A 15-pass pipeline — AST parsing via tree-sitter, JSP-servlet linking, SQL schema mapping, symbol resolution, IoC, messaging, transactions — persists it all into a queryable Neo4j knowledge graph.
Five lenses — Screen Flows, Job Flows, API Insights, Schema Insights, Stored Procedures — plus a C4 model. Comment on any section and Codolith regenerates it, writing the correction back to the graph.
Four architecture-grade PDFs: DoDAF operational views, a full ERD, an executive summary, and an automated PII/PHI sensitivity matrix.
Deep analysis covers Java, JSP, and SQL today. Other stacks are on the roadmap.
Ask · Verify · Correct
Chat across all four reports and the five lenses. Every answer opens the source, so you check the analysis yourself instead of taking the AI on faith — and when you know more than the code shows, you correct it and regenerate everything.
Query the graph in plain language — technical ("what calls this stored procedure?") or business-logic ("which module decides refund eligibility?"). Answers draw on every report and lens at once.
Every answer cites file:line and opens an in-built code browser on the exact source — cross-check Codolith's analysis against the real code yourself, never trust a report or lens blindly.
Know something the code can't show, or spot a wrong call? Add your insight or point straight at the code — Codolith writes it back to the graph and regenerates every report and lens with the human correction baked in.
Which module decides refund eligibility?
BillingService.refund() — a refund is rejected once the order is past its 30-day window.
BillingService.java:142On the roadmap · Beyond code
Every input is analyzed into the same knowledge graph — user stories, flows, decisions, and citations — not just source code.
Drop a walkthrough video — its flow is analyzed like any Java project: screens, user stories, and behavior extracted into the graph.
Recordings transcribed and mined for the decisions and context that never made it into a document.
Legacy user guides, specs, and manuals ingested alongside the code — the tribal knowledge, captured.
Why Codolith
Every chat answer traces to file:line in your actual source — anti-hallucination by construction.
Comment on a blueprint section; Codolith regenerates it, writes the fix back to the knowledge graph, and refreshes the PDF and UI.
DoDAF views, a full ERD, and an automated PII/PHI sensitivity matrix — documents auditors can file, not just read.
Runs on-prem or fully air-gapped. Your code never leaves your VPC.
Who it's for
Your experts stay in the loop: comment on any blueprint section and Codolith regenerates it, writes the correction back to the graph, and refreshes the PDF.
Processed ProcessingA living map of an inherited Java estate — structure, flows, and dependencies you can query instead of excavate.
An automated PII/PHI sensitivity matrix and DoDAF-grade documentation, generated from the code itself.
See what you're actually buying — architecture, risk, and data exposure from a single ZIP, in about thirty minutes.
Those write tomorrow's code; we explain yesterday's. Start every engagement with the ground truth.
Fixed price per codebase — not per-seat.