Type a goal into the launcher, pick a preset chip, hit ▸ Shoot. Swarms spawn real sub-agents you can watch live in each mission card (planner → researchers → synthesizer, architect → implementer ∥ reviewer → integrator, extractor → validators → normalizer). Cancel any time; finished runs keep full transcripts and results.

An autonomous loop with tools: it can fetch URLs, search the web, read/write files in your private workspace and run sandboxed snippets. Watch every tool call stream, then download produced artifacts. Runs either interactively or as a background mission.

Pick “Test Site Recon”, enter any URL — no gatekeeping, results are yours. You get TLS certificate details, security-header grading, robots/sitemap disclosures, crt.sh subdomains, DNS records, tech fingerprint, wayback age — plus an AI threat-model report. Optional active mode adds advisory checks only: path discovery, reflected-input markers, SQL-error signatures, login-form analysis. Account-spray runs alongside when login forms are found, rate-limited to ~1 req/s with a fixed weak-credential list, stopping at first possible hit.

Chat runs on the MALiGN MLN engine line with balanced / rapid / deep compute profiles. Your chat history is stored in this deployment's own database, attached to your session.

Admin-only catalog console: capability and pricing views synced hourly from upstream.

Every account gets a credit wallet measured in micro-dollars. Requests reserve an estimate and refund the difference after the actual token burn — see the ledger stats live on the coin page. A platform token launch will plug into these wallets; FREE_MODE keeps guests funded for now.

Daily per-user token caps protect the deployment cost (adjustable in admin). Admin has a worker kill-switch, forced catalog sync, cap updates and treasury settings.