Loading…
AgenticOp
Translate working software — with evidence, not guesswork.
Origin apps lift into WebIR, emit to real stacks, and prove behavior with oracle replay. This hub runs multi-site programs; the matrix below is what we can honestly claim today.
What we prove today
Live census from the hub capability matrix — fixture gold and product depth, kept separate on purpose.
-
627/627
Directed hub pairs at full fixture gold
Structured + middleware depth per origin→output cell (incl. COBOL origin file-lift). Trace replay in CI — not customer-repo parity.
-
72
Core language pairs with gold-fixture oracle
9×9 matrix (CWL, C#, Go, Java, JS, PHP, Python, Ruby, TS) with behavioral replay evidence.
-
24/24
WISP CWL outbound emit targets
Real management corpus (~106 routes) emits cleanly to every hub target language.
-
4
Priority deep prove (hono · fastify · nextjs · python)
Gold replay plus in-process WISP→hono/fastify correctness=1. Not D6448-ST on every emit.
-
ST
WISP product success template on CWL static
Evidence-only hole zero, signed-in origin compare, live management demo — the customer-depth bar.
Loading live matrix…
Where depth is real
Flagship and product work concentrates on backends we can verify end-to-end.
-
PHP
→ Hono, Fastify, Next.js, TypeScript
tiny-blog oracle micro + plain PHP / Symfony pilots with verify replay.
-
Node
Express → Hono
Live Express flagship with capture and oracle verify.
-
WISP
SvelteKit → CWL static (+ Firebase)
Full-site convert with evidence-only ST — the UI product proof.
How conversion runs
Translate-only method — no demo façades, no invented widgets.
- InventoryRoutes, adapters, gaps from the real origin.
- LiftOrigin idioms → WebIR; holes stay honest.
- EmitTarget stack handlers from WebIR, not stubs.
- ProveOracle replay, origin compare, operator evidence.
Emit targets
All 24 hub emits are proven outbound from WISP CWL. Highlighted names have selective deep prove.
- c
- cpp
- csharp
- css
- cwl
- fastify
- go
- hono
- html
- java
- json
- kotlin
- markdown
- nextjs
- php
- python
- ruby
- rust
- scala
- scss
- sql
- swift
- vue
- yaml
Honest limits
Do not read “627/627” as production-ready for every language. Fixture gold means the hub cell lifts, emits, and replays on gold suites. Customer cutover depth today is PHP (+ Express) backends and the WISP CWL product track. Ask for a pilot when you need your corpus on that bar — hello@agenticop.io.
Create your demo account
Sign in to use the public demo hub. Full pilots are uncapped — hello@agenticop.io.
Projects & workspace
Run this hub server from your browser: attach legacy sites over SSH, pick one web output language, then Run all sites with parallel jobs in the project Console.
Capabilities · Documentation · Migration OS, CLI, install, and operator library.
Organizations
Team projects share an org; members see org-scoped projects only.
Your projects
Hub stores workspaces under ~/.chrysalis-hub/ on this server.
Language completion tracker
Prioritized by popularity; shows what remains per language.
| Origin | Output | Grade | Path | Next action |
|---|---|---|---|---|
| — | ||||
Extended matrix oracle census
627/627 oracle-product (G9160 maintenance). Trace replay in CI — not production cutover parity.
Translation path explorer
Browse the 627-pair hub matrix: per-pair ingest/emit/verify lanes, similarities, differences, and consolidated cross-language synthesis.
Synthesis overview
Loading synthesis…
Gold pairs
—
Pair detail
Migration plan
Combine language compare, pair risk, and detected datastores into operator steps.
Similarities
Differences
Best practices
—
New translation project
1 Project & sites → 2 Auto setup (SSH) → 3 Console: translate all sites
Origin sites (SSH) — add as many as you need
Fill one site below and click Add to project. Repeat for each legacy server. All sites share the same output language.
0 sites queued
No server to point at? Upload your files
Skip SSH — upload your site's PHP files (or a .zip of the whole tree) straight from your browser.
You'll also need a small chrysalis.routes.json in the upload declaring your routes explicitly
(no auto-detection — Chrysalis never guesses routes from code).
Show an example chrysalis.routes.json
{
"app": "my-site",
"routes": [
{ "method": "GET", "path": "/posts", "file": "pages/posts_list.php", "pathParams": [] },
{ "method": "GET", "path": "/posts/:id", "file": "pages/posts_view.php",
"pathParams": [{ "name": "id", "type": "int", "phpVar": "post_id" }] }
]
}
Output language (all sites)
Set manually if autodetect is off or wrong.
SSH: add sites above, or use Local workspace only for code already on this hub. No CLI required.
Documentation
Chrysalis operator library — Migration OS, CLI, hub, governance, and WISP programs.
Console
Client/server hub — add many SSH origin sites, run in parallel with per-site progress.
Project settings
Translation route plan
Loading plan…
Migration plan (project)
Load a project to see migration steps.
Evidence dashboard
Verify correctness, holes, and blockers for the first site.
Origin sites
Batch: idle
Add site (SSH)
idle
Use Run full pipeline for prep + pull + translate in one click.
0%
0 / 0 routes (aggregate)
IS routing: idle
Hole closure: verify-gated repair bridge when holeId present
Oracle verify (staging traces)
Copy traces into each site’s .chrysalis/traces/ on this hub, start the emitted app, then verify.
Observe on staging
Select a site and load capture instructions (PHP oracle on staging, not on prod).
—