MigrationProof
MigrationProof helps developers modernize legacy applications and migrate workloads from one technology stack to another—with evidence that critical behavior is preserved.
Core trick
Independent verifier runs checks on each migrated candidate and returns failures to a persistent managed agent for repair before the next stage.
Tags
Description
MigrationProof helps developers modernize legacy applications and migrate workloads from one technology stack to another—with evidence that critical behavior is preserved.
Code conversion is only part of migration. The harder problem is identifying what broke, repairing it, and knowing whether the result meets the application’s requirements. MigrationProof combines Gemini 4 Argon’s code analysis and repair capabilities with Managed Agents’ persistent workspace and file-editing tools. An independent verifier checks each candidate and sends failures back to the agent for repair before allowing progression.
Our interactive 3D demo illustrates this workflow through a GPU-to-TPU migration candidate. A deliberately injected normalization fault stops the migration; Argon repairs it, recovering two failed checks and passing all 18 source-contract checks. The demo replays actual agent evidence, while the final hardware handover remains a preview pending real GPU/TPU validation.
GPU-to-TPU is one example of our broader vision: migrations across languages, frameworks, runtimes, and infrastructure, supported by workload-specific checks. MigrationProof makes changes, failures, repairs, and verification results visible—giving developers a reviewable basis for deciding whether a migration is ready to advance.
README excerpt
# MigrationProof MigrationProof is an interactive workbench for AI-assisted application and workload migrations. Gemini 4 Argon and Managed Agents generate and repair migration candidates; independent checks make changes, failures, and repair results inspectable. The included examples cover a CommonJS-to-TypeScript inventory importer and source preparation for a GPU-to-TPU embedding worker. The sample inventory importer is an example workload. The product is the migration, verification, repair, and evidence workflow. ## Run locally Requires Node.js 24 or newer and a Gemini API key with access to `gemini-4-argon` and `antigravity-preview-09-2026`. The accelerator source checker additionally requires Python 3.10+ (`python` on PATH or the `PYTHON` environment variable). 1. Run `npm install`. 2. For live agent calls, copy `.env.example` to `.env` and set `GEMINI_API_KEY` there. The prepared 60-second demo and seeded CSV example work without a key. The key is used only by the Node server. `.env` is ignored by Git. 3. Run `npm test`. 4. Run `npm run build`. 5. Run `npm start`, then open [http://localhost:8787](http://localhost:8787). The server binds to `127.0.0.1`. For UI development, run `npm start` and `npm run dev` in separate terminals, then open [http://localhost:5173](http://localhost:5173). ## Demo paths - **Prepared 3D migration** at [http://localhost:8787/demo](http://localhost:8787/demo): a 60-second replay of the verified real preparation and repair pair, with c