The Platform · The layer underneath

A substrate, and the specialists that plug into it.

Most consulting firms sell frameworks and talent. We sell those too — and a collection of software we've built to run the frameworks. Every engagement ships with the Platform, under universal license, customized for your stack.

Architecture Substrate-plus-specialist
Tools in active use A dozen, across three layers
License model Universal, time-bounded, artifacts yours

Three layers. Every tool lives in one of them.

Our Platform isn't a suite. It's an architecture. Three tools run across every engagement regardless of lane, regardless of phase — the substrate. A set of specialists plugs in at specific points of the AI-First SDLC. A smaller set activates only inside specific engagement shapes.

This shape isn't cosmetic. The substrate is what lets specialists compound. Every downstream tool reads from the same context, writes into the same workflow, and inherits the same ingestion layer. A spec drafted in Nova flows into Cycle, informs Vulcan's next ticket, and gets reviewed against quality profiles stored in Orbital. A decision captured by Ephemeris in a meeting surfaces the next morning as a pending work item.

We built it this way because the alternative — specialists with private memory — doesn't compound. It fragments.

The three tools that run across every engagement.

Orbital

The context layer.

Project documentation repository — ADRs, PRDs, planning docs, research, interview notes, decisions — stored as embeddings with a conversational interface. Every downstream tool reads from Orbital. When a new engineer, a new agent, or a new stakeholder needs context, they query Orbital rather than interrupting someone.

Role · Context-as-infrastructure
Cycle

The workflow layer.

AI-First SDLC workflow application. Cycle and sprint planning, team metrics, verification workflows, stakeholder governance. Replaces JIRA-shaped tooling with spec-first, evidence-first work units. Every specialist that produces output — a spec, a resolved ticket, a rebuild increment, a parity report — lands in Cycle.

Role · Work-unit orchestration
Ephemeris

The ingestion layer.

Meeting processor. Runs on a schedule, converts transcripts into structured summary notes, and routes decisions, work items, and action items into Cycle while routing insights and reference material into Orbital — all after human review. Meetings stop leaking.

Role · Everything-else-to-substrate ingestion

Six phases. Each has its specialists.

The daily rhythm of AI-First delivery. Each phase has tools that specialize in it — substrate tools span all of them, phase specialists focus where their leverage is highest. This is how an engagement actually runs, day to day, week to week.

Phase 01
Discovery
Surfacing the problem space. Stakeholder interviews, research, prior art, constraints. The output is context — captured in a form other tools can read.
Tools at work
EphemerisInterview transcripts → structured notes → Orbital.
OrbitalResearch, interviews, and decisions stored with conversational query.
Phase 02
Planning
Scoping into cycles. Pulling decisions out of Discovery, shaping work units, sequencing. The output is a sprint-shaped set of work items with verification criteria already attached.
Tools at work
CycleCycle planning, stakeholder governance, work unit shaping.
OrbitalContext fed into planning conversations.
Phase 03
Specs
The leverage point. A well-drafted spec is a document agents can act on and reviewers can evaluate — one that passes testability checks has its failure modes already named. In AI-First work, tech design often collapses here: the spec carries enough to begin, and the design emerges inside the loop.
Tools at work
NovaDrafts specs from templates, verifies testability, generates epics into Cycle.
OrbitalPrior-art lookup, pattern references, ADR traceability.
Phase 04
Coding
Where the Platform has the most surface area. Agents write code against specs, run tests, open PRs, and submit to adversarial review. Humans review, approve, and land.
Tools at work
VulcanAgentic ticket resolution — plans, executes, tests, opens PRs.
DebrisMaintenance work — refactors, upgrades, cruft reduction.
ParallaxMulti-approach exploration — parallel agents try different solutions.
QuorumMulti-model review council — adversarial verification before human review.
Phase 05
Deployment
Landing work into production. The Platform is deliberately light here — your CI/CD, your environments, your release rituals. We integrate; we don't replace. This is a place where your stack stays yours.
Tools at work
CycleRelease tracking, work unit state, deployment gates.
Phase 06
Operations
Production reality. Incident response, observability, routine maintenance. Lane specialists activate here when engagement shape demands it — Lineage for AI code-path audits, Atlas for agent inventory, Gemini for ongoing parity.
Tools at work
CycleIncident and ops work units.
OrbitalPostmortems, runbooks, ops decisions.
Lane specialistsActivate where engagement shape demands.

Built for specific parts of the loop.

These are the tools that run the day-to-day AI-First delivery loop. They read from the substrate, write into it, and handle the hard parts of spec authorship, agentic work, exploration, and adversarial verification.

Nova

Spec authoring & testability

Drafts specs from templates, verifies testability, auto-generates epics into Cycle. The leverage point for the whole practice — the document everything else pivots on.

Vulcan

Agentic ticket resolution

Browses code, plans, executes, tests, opens a PR for human review. The production-grade "agent fixes a ticket" loop your team can actually trust.

Quorum

Multi-model review council

Several LLMs review changes in parallel and adjudicate each other's findings. Only the surviving items reach the human reviewer — review latency down, review signal up.

Debris

Maintenance agent

Handles the refactors, upgrades, and cruft-reduction work that accumulates silently in any codebase. Ships small, reviewable PRs on a cadence.

Parallax

Multi-approach exploration

For problems with several viable directions, Parallax runs parallel agents trying different approaches, then surfaces the strongest for human selection. Exploration as a first-class activity.

Ephemeris

Meeting ingestion

Listed here as well because it's the specialist that keeps the substrate honest. If it didn't route meeting output into Cycle and Orbital, those layers would starve.

Phase specialists run inside AI Transformation engagements by default, but they're not exclusive to that lane — a Modernization rebuild still uses Vulcan for ticket work, still uses Quorum for review. The lane determines which specialists activate most visibly, not which are available.

Activated inside specific engagement shapes.

Tools whose reason to exist is a specific class of engagement. They read from the substrate, write into the workflow, and handle work that only shows up when the engagement shape demands it.

For Modernization — Legacy rebuilds & capability transfer
Strata

Legacy system analyzer

Per-module analysis with an adversarial writer/critic loop. Produces a seam map grounded in per-module evidence — turning archaeology into validation.

Gemini

Parallel-rebuild evidence

Continuous parity between BAU and rebuild systems. Makes the "are these actually equivalent?" question a dashboard, not a debate.

Transit

Capability-transfer tracking

Milestone-based assessments of receiving-team AI-First readiness. Makes capability transfer a measurable handoff, not a hopeful one.

For Security — Defensive AI capability
Orion

Two-agent agentic pen-test

A hunter agent finds vulnerabilities; a resolver agent proposes fixes. Uses the same class of models and patterns real adversaries use.

Atlas

Agent inventory & audit

Discovery and classification of agents-in-production, their scopes, and their audit-trail shape. The map of your agentic attack surface.

Lineage

AI code-path inventory

Identifies AI-generated code regions, their review state, and their risk profile. The map of your AI-written code, by risk.

Lane specialists don't run in every engagement — they activate where the engagement shape demands them. A Transformation engagement without Modernization scope won't use Strata or Gemini; a pure delivery engagement without Security scope won't use Orion or Atlas. Every engagement, regardless of lane, runs on the substrate.

Universal license. Artifacts stay yours.

Every engagement includes a universal, time-bounded license to every Platform tool relevant to the work. Every relevant tool, customized for your stack and supported by us for the full duration of the engagement. No per-seat metering, no per-feature gating.

When the engagement winds down, the license winds down with it — but the artifacts the Platform produced stay with you. The spec library in Orbital, the cycle history in Cycle, the posture report from Orion, the parity evidence from Gemini, the capability-transfer record from Transit. These land in your repositories, owned by your team, portable to whatever stack you run next.

The Platform is how our engagements compound. You're not renting tooling; you're installing an architecture that outlasts us.

Ask what ships with your engagement.

Start a conversation Browse the service lanes Direct to a principal. No SDR filter.