Canonical ICP lives elsewhere. The "who is this for" question for AINL as a whole is answered in
../WHO_IS_THIS_FOR.md— this document is ArmaraOS-specific product positioning: the stack diagram, install paths, persona-to-message mapping, and reference Hands. Read the canonical ICP doc first if you're trying to decide whether AINL fits your workload; come here for the product angle once you've decided AINL is in the running.
Many mature teams already run deterministic scripts + LLM at judgment gates. For them, pitching "AINL saves 80% vs your cron job" fails — see ../WHO_IS_THIS_FOR.md for the full anti-pitch.
ArmaraOS is the primary product wedge: packaged agent OS + dashboard + validated Hands — not "replace your bash monitor with a compiler."
ArmaraOS runs agents; AINL compiles the deterministic work they repeat. Download once, schedule validated workflows, audit executions — without re-prompting the model every cron tick.
The full persona table — solo operators, agent-heavy shops, small teams needing audit, Temporal-curious teams, compliance-led shops, researchers — and the "not primary ICP" boundary live in ../WHO_IS_THIS_FOR.md. What this doc adds on top of that is the product mapping below: which ArmaraOS capability each persona reaches for first.
┌─────────────────────────────────────────────────────────┐
│ ArmaraOS desktop / daemon (~32 MB) │
│ Dashboard · chat · cron · MCP · graph memory │
└───────────────────────────┬─────────────────────────────┘
│
┌───────────────────┼───────────────────┐
▼ ▼ ▼
ainl-mcp scheduled App Store
(authoring) ainl run (Hands)
│ │ │
└───────────────────┴───────────────────┘
│
▼
AI_Native_Lang compiler + runtime
(strict IR · adapters · audit JSONL)
- Hands (openfang-hands packages) —
ainl emit --target armaraosproducesHAND.toml, IR JSON,security.json. Operators install from App Store; kernel schedules runs. - MCP authoring loop —
ainl_get_started→ainl_validate→ainl_compile→ainl_runwithwizard_state_json. Agents compile before they execute side effects. - Scheduled deterministic execution — kernel injects
AINL_ALLOW_IR_DECLARED_ADAPTERS=1on cron children;bundle.ainlbundlefor graph-memory round-trip. - Graph memory + persona — Rust
ainl_memory.dbper agent; extractor/tagger optional; pairs with proceduralmemory.patchon Python side. - Dashboard observability — usage, eco mode, graph-memory prompt accounting, orchestration traces.
De-emphasize in headline GTM: 67 adapters, Solana, TikTok, blockchain — document for integrators, not homepage hero.
| Program | Role | LLM at runtime |
|---|---|---|
examples/benchmark/enterprise_monitor.ainl |
HTTP health monitor | 0 when healthy; 1 on incident |
examples/workflows/support_ticket_router.ainl |
Triage + draft | 3 content calls; 0 routing tokens |
examples/workflows/data_pipeline.ainl |
Multi-adapter ETL | Deterministic + adapter I/O |
openclaw/bridge/wrappers/token_budget_alert.ainl |
OpenClaw daily budget digest | 0 (bridge adapters only) |
| Message | Use when |
|---|---|
| "Download ArmaraOS — agents + dashboard + Hands" | Consumer / prosumer landing (ainativelang.com) |
"Add ainl-mcp to your existing agent" |
OpenClaw / Cursor / Claude Code integrators |
| "90–95% vs prompt-loop agents on recurring jobs" | Teams still re-prompting every cron (baseline A) |
| "Compile-once audit trail + strict validation" | Compliance, agent-authored ops |
| Avoid for baseline B/C teams |
# Desktop product (recommended for non-developers)
# https://ainativelang.com/armaraos
# Existing MCP agent
pipx install 'ainativelang[mcp]' && ainl setup --auto
# ArmaraOS daemon MCP merge
ainl install-mcp --host armaraosHub docs: docs/ARMARAOS_INTEGRATION.md, armaraos/docs/scheduled-ainl.md, armaraos/docs/mcp-a2a.md.
../WHO_IS_THIS_FOR.md— canonical ICP and anti-ICPPRODUCTION_EVIDENCE.mdFROM_LANGGRAPH_TO_AINL.mdAINL_AND_TEMPORAL.md