Capability tiers
Every website gets a URL audit. Source mapping and automatic fixes are capability-gated, and every output names the tier that applied.
| Tier | Capability | Allowed claims |
|---|---|---|
| T0 | HTTP, redirects, robots, sitemaps, metadata, structured data, server HTML, internal links | Observed URL findings; no repository provenance or patches |
| T1 | Non-executing repository inventory, framework hints, candidate files | candidate/unmapped only; no automatic patch |
| T2 | Framework-aware route and metadata resolution | File and line only when the evidence meets the provenance bar |
| T3 | Closed-list patches with approval, snapshots, rollback and verification | Only listed fixes, only when every precondition holds |
| Agent | Contextual investigation using T0/T1/T2 evidence | Agent judgement, never engine provenance or score |
Platform matrix
This is a capability roadmap, not a claim that every integration has been tested. Detection is a hint about where source lives; it is never proof of how runtime output is produced.
| Platform | URL audit | Repo / agent | Source provenance | Patches |
|---|---|---|---|---|
| Next.js App Router | T0 | T1 + agent | T2 (this release) | T3, closed list |
| Static HTML | T0 | T1 + agent | T2 (this release) | T3: S1, S2, G1 |
| Next.js Pages Router | T0 | T1 + agent | Not in 1.0 | None |
| React + Vite / SPA | T0; renderer recommended | T1 + agent | Not in 1.0 | None |
| Vue / Nuxt | T0 | T1 + agent | Post-1.0 candidate | None in 1.0 |
| Angular | T0 | T1 + agent | Post-1.0 candidate | None in 1.0 |
| Svelte / SvelteKit | T0 | T1 + agent | Post-1.0 candidate | None in 1.0 |
| Astro | T0 | T1 + agent | First post-1.0 candidate | None in 1.0 |
| Remix / React Router | T0 | T1 + agent | Post-1.0 candidate | None in 1.0 |
| Gatsby / Eleventy / Hugo / Jekyll | T0 | T1 + agent | Post-1.0 candidate | None in 1.0 |
| WordPress / Drupal / Shopify / Webflow / Wix | T0, where reachable | Agent, if source is available | Not in 1.0 | None |
| Other server-rendered stacks | T0 | T1 + agent | Not in 1.0 | None |
Framework detection
Detection reads package.json, lockfile names, configuration files and directory shapes as text. It never runs repository scripts or configuration. It reports detected, suspected or unknown with its evidence; conflicting signals produce suspected or unknown.