Product alignment — which component ships in which product

Mapping the architecture to the three commercial products (Runtime / Control / Intelligence). Every component lives in one of these products. This is the investor / pricing / go-to-market view.

The three products from Strategy Memo v3 §11:

Product	Buyer	Motion	Pricing	Status
Runtime (open-core)	Developer	Viral, `pip install`	Free forever	MCS Q2 2026
Control (SaaS + on-prem)	CISO / Head of AI / GRC	Enterprise sale	Per governed agent / month	Q3-Q4 2026
Intelligence (data moat)	Existing Control customer	Upsell	Telemetry volume bands	2027

Component → product map

Runtime (open-core SDK + CLI + providers)

The free, viral, developer-installed product. Everything required to run a governed agent on a single machine without a server-side dashboard.

Component	Why Runtime	Status
Compiled Policy schema	Canonical IR; ships with the SDK	concept (Q3)
Policy compiler	Compiles Compiled Policy from Policy + sandbox-spec; runs in Runtime mode against local Policy file	concept (Q3)
Cascade merger	Merges org/project/agent levels; degenerates to single-level for solo Runtime use	concept (Q3)
Live policy push channel	Listens for updates; works against either local Policy file or remote Control plane	concept (Q3)
Operator CLI `tappass`	Authoring + provisioning surface; works locally or against Control plane	concept (Q3)
Host runtime CLI `tappass-host`	Receives sync; applies layered config; launches agent	concept (Q3)
Agent client SDK `tappass-agent`	The library every agent imports	concept (Q3)
Upstream tool proxy / MCP forward	MCP-side enforcement; ships with Runtime	concept (Q3)
Resource access checker	Pipeline step; ships with Runtime	concept (Q3)
Runaway agent stopper	Pipeline step; ships with Runtime	concept (Q3)
Approved tool-server list	Per-org MCP registry; works in Runtime mode against local config	concept (Q3)
Kernel ring applier	OpenShell + Landlock + L7 network	partial (OpenShell shipped)
Harness ring applier	settings.json writer + per-runtime adapters	concept (Q3)
Interpreter ring applier	Monty / V8 / Wasmtime profiles	concept (Q4)
LLM gateway (cross-cutting)	OAI / Anthropic / MCP / LiteLLM	✅ shipped
OpenShell sandbox + nono	Kernel-ring primitives	✅ shipped
32-step pipeline	Detection / capability tokens / audit	✅ shipped
Hash-chain audit	Integrity verification	✅ shipped
ES256 mandates	Per-allow signing	✅ shipped
Authoring resolver (intent-to-policy)	Function/category/concern/capability collapse to pipeline	✅ shipped on main

Runtime promise: every part of governance up to and including local enforcement works without a Control plane. Rego policy file → 5 enforcement positions → agent runs.

Control (SaaS + on-prem dashboard)

The commercial product. CISO / GRC pays for it. Adds central authoring, multi-machine sync, audit dashboards, marketplace, compliance reports.

Component	Why Control	Status
Onboarding wizard	Click-driven Policy authoring for non-engineers	concept (Q4)
EU AI Act compliance pack	Procurement-defensible; certified once, inherited	concept (Q3)
OWASP LLM Top 10 pack	Same	concept (Q3)
GDPR / PCI-DSS / HIPAA / NIS2 / DORA packs	Same	planned (2027)
Sandbox registry / state store	Multi-machine "where are my agents?" view	concept (Q3)
Reconciler	Drift detection across the fleet	concept (Q3)
Audit dashboard	Visual trace timelines, denial reasons, replay	concept (Q3)
Authoring UX (GitOps, simulation, shadow mode)	Strategy memo §09 vector 05	concept (Q3-Q4)
Marketplace v1 (3 certified policy packs)	Switching-cost lock-in	concept (Q4)
Pre-deployment evaluator	Runs probe suite against agent before deploy; CI integration	concept (Q4)
OWASP LLM probe library	Probe content for the evaluator	concept (Q4)
SOC 2 Type 1 audit	Procurement filter for enterprise / EU buyers	calendar (Q3-Q4)
Public trust page	0-training, residency, encryption, sub-processors	concept (Q3)

Control promise: one Policy authored centrally, applied everywhere your agents run, audited in one dashboard, with auditor-ready compliance reports out of the box.

Intelligence (data-moat upsell)

The 2027 product. Cross-customer behavioral intelligence. Only valuable once Runtime + Control reach scale.

Component	Why Intelligence	Status
Behavior drift monitor	Production behavior vs. baseline; cross-customer pattern detection	concept (Q4 single-tenant; Intelligence layer 2027)
Cross-customer anomaly detection	"Your agent is the only one in our network using this tool pattern"	planned (2027)
Industry benchmarks	"Peer fintechs deny this tool by default"	planned (2027)
MCP / skill vulnerability disclosure	Push policy update within minutes of disclosed flaw	planned (2027)
Behavioral baselines per industry	Compounding from telemetry scale	planned (2027)

Intelligence promise: the security signal no single customer can build alone. The longer we run, the sharper our cross-customer signal becomes.

What this map enables

Pricing conversations

Runtime is free. Control is the pricing surface. Intelligence is upsell.

A buyer asking "what does this cost?" gets:

Runtime: $0
Control: per governed agent / month, tiered, on-prem surcharge
Intelligence: telemetry-band add-on, opt-in, available 2027

Investor conversations

Three flywheels, one substrate:

        Runtime (free, viral)
                │
                │ every install = lead for Control
                │ every install = telemetry for Intelligence
                ▼
        Control (per-agent SaaS)
                │
                │ scale enables
                ▼
        Intelligence (cross-customer signal)
                │
                │ retention + pricing leverage
                ▼
        compounding moat

Engineering prioritization

Within each product, what to ship first:

Runtime Q2 2026 critical path = Compiled Policy + 2 providers across 2 rings + push/pull/reconcile + state store. The MCS. Nothing else gets prioritized over this.
Control Q3-Q4 2026 critical path = Authoring UX + audit dashboard + SOC 2 Type 1 + 2 compliance packs.
Intelligence 2027 = single-tenant drift first; cross-tenant after Q4.

Sales motion

Buyer	First contact via	Conversion path
Solo developer	`pip install tappass`	→ joins team → buys Control
Mid-market CTO	Compatibility matrix → Control trial	→ SOC 2 close → Control contract
Enterprise CISO	Compliance pack → on-prem Control demo	→ procurement → on-prem Control + Intelligence opt-in
Vendor (e.g. Collibra)	Reference architecture → Control resale / partnership	→ vendor offers TapPass-governed runtime to their customers

Pricing strategic principles

Never gate core enforcement. Runtime ships every enforcement position; Control adds the operator/CISO experience.
Open-core distribution flywheel. Free Runtime is what makes a paid Control competitor unable to catch up (no flywheel).
Charge for outcomes, not features. Control's pricing is per-agent because that's what scales with customer value.

What this means for the roadmap

The three products give the roadmap a coherent strategic spine:

Quarter	Product focus
Q2 2026	Runtime MCS — the four-piece v1 deliverable
Q3 2026	Runtime extended (more providers) + Control v1 (authoring, dashboard, SOC 2 prep)
Q4 2026	Control v1 GA (compliance packs, marketplace, eval, drift) + MCP broker + chat-bot surfaces
Q1 2027	Intelligence alpha + Federation + Compliance v2

See roadmap/2026-h2.md for week-level detail.

References

Strategy Memo v3 §11 — Product shape
Strategy Memo v3 §14 — Moats and defensibility
OVERVIEW.md — three products in the 1-pager
COMPATIBILITY-MATRIX.md — what each runtime delivers
concepts/minimum-credible-substrate.md — what gates Runtime v1