© 2026 LyDos · v12.3.0-beta
Compliance-grade execution for engineering teams. Every command runs on real machines with cryptographic proof — your data stays where you put it, your audits write themselves.
Verified Sprint 12 comprehensive audit. Numbers regenerate from manifest.
Install once. Works on Linux, macOS, and Windows. Then run anything.
curl -fsSL https://lydos.ailydian.com/install | shAll 7 AI CLIs work with `lydos login`
Vendor-locked assistants want you to switch. Solo CLIs want you to juggle. LyDos coordinates the seven major AI CLIs you already use, adds a private engineering corps trained against your project, and signs every action with a key only your device holds.
Claude Code, Cursor, Codex, Aider, Cline, VS Code and Windsurf — running through one shared session pty, one shared workspace, one shared audit chain. You don't pick a CLI; LyDos picks the right one for the step.
Read the runtime docsAn autonomous engineering corps boots inside your org_id-scoped sandbox the moment you sign in. They are not chatbots — they execute, sign, and audit on your behalf, against your tree only. Your code never leaves your sandbox.
Browse the corpsEvery CLI session spawns inside q243 IsolatedEnvironment with your workspace pre-mounted, your secrets pre-loaded from the per-user vault, and your audit policy pre-applied. You type the goal; LyDos provides everything else.
Read the architecturestdin and stdout are never written to the database — only their SHA-256 hash, in a hash-chained audit table. Your secrets are encrypted per-user with a key derived from your master + a per-user salt. The CLI binary itself runs inside namespace_jail + seccomp + cgroup.
View Trust CenterA real pty stream over WebSocket — not a chat transcript. Resume on your phone what you started on your desktop, with a 4-hour session TTL and a mobile data toggle that drops bandwidth from 64 MiB to 1 MiB on cellular.
Open mobile shellEvery critical action passes through a layered defence: hardware root of trust, key sovereignty, settlement integrity, autonomous response, and pattern intelligence.
Each engine ships its own report (docs/*_REPORT.md) + route prefix.
No marketing inflation — these are the 50 production classes in core/q158_*.py on the live VPS. Each engineer scans, observes its result, and learns continuously through the RepairIntelligenceEngine (4-layer MCTS+UCT patch search + confidence gate + observed-pattern memory).
Checks API endpoints, health routes, response codes
Backend container memory + error log patterns
Cloudflare proxy status + cache headers
Outdated dependencies + known vulnerabilities
Container health, restart counts, resource usage
.env safety, hardcoded secrets in config
Frontend assets, broken images, favicon
Repo dirty state, .gitignore, large files, secrets
ruff (Python) + ESLint (JS/TS) runs
Response times, TTFB, page weight
Viewport meta + mobile responsiveness
Meta tags, OG, structured data, robots, sitemap
Security headers, SSL, exposed paths, server disclosure
Deployed-domain header + technology-leak audit
Test existence + test runner + script validation
tsc compilation, strict mode, type coverage
Chief Engineering Officer — orchestrates all engineers
Default-credential detection
Hidden path + open directory discovery
SQLi / XSS / LFI / XXE payload fuzz
Source-code sensitive-data leak scan
Root-cause from Docker logs + service dependency chain
Service resilience + chaos-test posture
PostgreSQL perf analysis + tuning hints
Last-deploy + image-age + health verification gate
Feature flag inventory + stale flag detection
MTTR calculation + engineer effectiveness
Commit msg quality, test presence, branch hygiene
Container resource right-sizing recommendations
Python + JS/TS static analysis with auto-fix
High-entropy strings + hardcoded secrets
Code review on recent diffs (NIM code models)
Documentation-quality analysis (NIM LLM)
Embedding search for issue-pattern code
Whole-codebase reasoning on Nemotron-3-Super-120B
Screenshot + document image OCR
PII detection + masking in configs
Security-issue scan via NIM LLM
True semantic code duplication (Llama-Nemotron)
WCAG accessibility audit via HTML analysis
CI/CD pipeline + last commit + deploy script
Container resource use + cost optimisation
PG/MySQL health, connections, large-table watch
i18n config + missing locale + lang-attr
Docker log analysis — spike + memory-leak detection
SSL expiry, TLS version, HSTS preload
Multi-endpoint uptime + response time
Multi-turn fix loop (scan → plan → act → observe)
Outside-world research feed (CVE / GHSA / NVD / OSV)
Opt-in OSINT scraping with sovereign gates
Per-domain global SEO/GEO keyword + page strategy
51 engineer classes verified by AST walk on production. When you sign in, every run is tagged with your tenant; only your members see your audit chain (PG RLS).
Architectural comparison
Every row is an architectural claim. Every cell is verified against the vendor's own docs as of 2026-05-18. The LYDOS column cites a source file in the repo.
| Capability | Composio | Zapier | n8n | Activepieces | LYDOS | LYDOS evidence |
|---|---|---|---|---|---|---|
| Per-user AES-256-GCM credential vault | core/integrations/credential_vault.py | |||||
| Per-call device-bound Ed25519 KSL signature | core/integrations/ksl_integration_gate.py (3-mode) | |||||
| 90-second-TTL per-request rotation key | core/integrations/request_rotator.py | |||||
| 5-surface unified vault (web/mobile/desktop/CLI/MCP) | core/integrations/cross_surface_vault.py | |||||
| Hash-chained audit (sha256[:16] only) | core/infrastructure/asr.py + KURAL 23 | |||||
| LSIA auto-quarantine (50 events / 10s → 60s 429) | core/integrations/immunity_engine.py (KURAL 22) | |||||
| 121+ integrations (each probed against real upstream) | core/integrations/credential_vault.py (121 entries) | |||||
| Self-hostable | Single-VPS docker compose today; HPA in FAZ D |
If a cell is wrong, open an issue with the source — this table is meant to be verifiable, not bold.
Real prices from the live `/api/payment/pricing` source. No surprises, no asterisks, no “contact for quote” paywalls outside Enterprise.
For trying things out
Per seat · individual & small teams
Customizable for organizations
Passkey-first · Magic link fallback · 15-min TTL · httpOnly · no password