reference
Intended Documentation
Platform Roadmap Programs
Honest status of Intended's strategic programs — what is implemented in code, what is partial, and what is Roadmap — with explicit promise boundaries.
Platform Roadmap Programs#
This page describes six strategic programs and their honest status against the code. It uses the same classification as the Capability Truth Matrix:
- Implemented — present in code and operational.
- Partial — a core surface exists; a documented refinement is not yet wired (the gap is named).
- Roadmap — designed or scaffolded, not active in the runtime today.
Warning
This is a program-status page, not a feature catalog. Depend on Implemented rows; treat Partial rows as "core only," and Roadmap rows as not yet available. Nothing here overrides the Capability Truth Matrix.
1. Multi-Step Workflow Engine#
What it does. Governs a chain of agent actions as one authorized workflow — each step independently authorized, audited, and fail-closed.
| Component | Status | Notes |
|---|---|---|
| Workflow API routes | Implemented | workflows routes; CLI workflow-run / workflow-status / workflow-resume / workflow-intervene |
| Step-level authority evaluation | Implemented | Each step passes the authority gate |
| Workflow database schema | Implemented | Persisted run/step state |
| Pause / resume / intervene | Implemented | workflow-intervene --action pause or cancel; workflow-resume --approvals … |
| Dynamic workflow planner | Roadmap | Not evidenced as a runtime planner today |
| Rollback / compensation | Roadmap | Compensation semantics not wired in the runtime |
Promise boundary. Workflows do not execute a step without per-step authorization. We will not ship a workflow path that bypasses the authority gate.
2. Connector Expansion#
What it does. Extends authority enforcement to downstream systems through the connector SDK and the tenant connector catalog.
| Component | Status | Notes |
|---|---|---|
Connector SDK (BaseAdapter) | Implemented | Fail-closed token validation before execution |
| Adapter registry | Implemented | Register and resolve adapters |
| Connector catalog (GitHub, Slack, Jira, ServiceNow, AWS, Azure, Salesforce, Datadog, Stripe) | Implemented | Present in the tenant connector catalog and connectors routes |
| Per-connector live execution depth | Partial | Catalog and fail-closed envelope are uniform; execution depth varies by connector. Confirm the specific connector before relying on a write action |
Promise boundary. Every connector validates an Authority Token before execution. No connector operates without authorization.
3. Intent Compiler & Domain LIM Ecosystem#
What it does. Compiles natural language into structured intent and supports installable domain LIM packs and a policy-pack marketplace.
| Component | Status | Notes |
|---|---|---|
| Intent compiler (LIM) | Implemented | NL→intent; ranked hypotheses, primary drives the decision |
| LLM backends | Implemented | rules, llm-api, local-model backends |
| Domain LIM packs | Implemented | lim-packs-list, lim-pack-install, installation lifecycle commands |
| Policy-pack marketplace | Implemented | marketplace/policy-packs with publishers, verification, provenance, and download |
| Full 173-category text classification | Roadmap | The text LIM classifies a small action set today; only the physical classifier emits OI-NNNN |
| Markov conformance scorer | Roadmap | Prototype, not wired; deterministic scoreSequenceConformance is the wired scorer |
| Federation protocol | Roadmap | Not implemented in the runtime |
Promise boundary. Marketplace packs are provenance-verified before installation; no pack bypasses authorization.
4. Billing & Commerce#
What it does. Usage-based metering and billing for platform consumption.
| Component | Status | Notes |
|---|---|---|
| Stripe billing integration | Implemented | stripe-billing and a usage meter reporter exist server-side |
| Usage metering | Partial | Metering plumbing exists; reconcilability against the audit trail is the design goal, not a verified guarantee here |
| Subscription lifecycle / plan entitlements | Partial | Present in part; confirm specifics before depending on entitlement enforcement at the gateway |
Promise boundary. Entitlement checks are intended to be fail-closed (unknown billing status denies). We will not silently allow overages, and metering is designed to be reconcilable against the audit trail.
5. Audit Evidence & SOC 2 Program Tooling#
What it does. Provides the tamper-evidence mechanisms and the program-management tooling an organization uses to run a SOC 2 engagement.
| Component | Status | Notes |
|---|---|---|
| Immutable audit hash chain | Implemented | Per-tenant SHA-256 AuditEntry chain with content + linkage verification |
| Signed evidence export | Implemented | HMAC-SHA256 keyed off the tenant secret (not publicly verifiable) |
| SOC 2 program management | Implemented | Engagements, evidence packs, control-evidence records, auditor requests, attestation handoff (/assurance/soc2/*) |
| Control-evidence capture | Implemented | Manual and auto evidence capture for tracked controls |
| Certified CC1–CC9 control mapping | Roadmap | The runtime audit chain encodes no control IDs; there is no built-in certified control mapping |
| External auditor attestation | External dependency | The attestation document is issued by an external auditor; Intended ingests and verifies it but does not issue it |
Warning
This program delivers evidence and workflow tooling, not a certification Intended can issue. Intended is not itself SOC 2-certified by virtue of this tooling, and the runtime does not map controls. Do not represent these capabilities as a built-in attestation.
Promise boundary. We do not fabricate external certifications. Auditor-issued attestations remain external artifacts.
6. Cross-Tenant Benchmarking#
What it does. Anonymized peer benchmarks for authorization performance.
| Component | Status | Notes |
|---|---|---|
| Per-tenant metrics | Implemented | Decision and request metrics are emitted (Prometheus) |
| Policy simulation | Implemented | Compare / drift / blast-radius |
| Cross-tenant aggregation & benchmark API | Roadmap | No dedicated benchmark API or dashboard route in the codebase today |
Promise boundary. When delivered, benchmarking will never expose tenant-identifiable data, will require explicit opt-in, will suppress small cohorts, and will not affect authorization decisions.
Related Resources#
- Capability Truth Matrix — the authoritative per-capability status
- Platform JTBD — persona workflows
- Go-Live Runbook — production rollout
- Architecture Overview — system components