Platform

How Foundry
builds.

Forge, Temper, Anvil, and Lore are products your team touches every day. Underneath each of them is a small shared platform — a workflow engine, a measurement loop, and an institutional memory system. This page shows how the layers fit together.

Forge
The cockpit. Renders workflows visually, surfaces AI agent actions in real time, and gives teams the control plane to supervise, approve, and intervene.
Temper
The measurement loop. Watches agent outcomes, detects quality drift, and runs controlled experiments to find what improves it.
Anvil
The workflow engine. State machines defined as data — the same engine drives software development, financial review, and operations workflows.
Lore
Institutional memory. Decisions, architectural rationale, historical traces, and prior art — stored in a form both people and AI agents can query.
Why we built it

Off-the-shelf AI tooling
can't encode your business.

We kept running into the same wall building for clients. The infrastructure didn't exist at the right shape, so we built it.

SaaS AI ships the same product to everyone.
Your business is not everyone.

The problem

Vendors ship one product to thousands. The AI never knows your specific business: your data, your terminology, the decisions your best people already made.

The longer story
SaaS AI tools can't capture institutional knowledge specific to one business. The AI you buy off the shelf doesn't know how your estimators talk about job risk, or what your ops lead decided three times before. Consultants leave a deck behind, not a system. The knowledge stays in a PowerPoint and the next problem starts from zero.
What we needed

A substrate that could capture knowledge, define workflows, run measurements, and keep everything alive past the engagement. At the right shape for mid-market businesses.

What that means in practice
Foundry needed a substrate where every engagement could do four things durably: capture knowledge, define workflows, run measurements, and keep the system alive past the engagement. That substrate didn't exist off the shelf.
What we built

So we built it. Not as a research project; as the engine underneath everything we ship. Every engagement runs on it. Every improvement compounds across every engagement that follows.

Why it compounds
Patterns that work in construction estimating inform how we structure financial review workflows. Measurement techniques proven in one vertical generalize to others. Every engagement raises the baseline for every customer that follows.
Generic SaaS Same product to everyone
+ Foundry engagement
Your kit Workflows, memory, measurement built for your business
each retainer cycle
Compounding Gets smarter with every engagement, automatically
GENERIC SAAS Same product to everyone Foundry engagement YOUR KIT Workflows, memory, measurement — yours each retainer cycle COMPOUNDING Gets smarter with every engagement, automatically
The four layers

Platform layers,
not separate products.

None of these are things you buy separately. They're the engineering foundation Foundry uses to build products that hold up — and that compound in capability every time we ship.

Anvil — Workflow Engine

State machines defined as data, not code. The same engine drives software development, financial review, and custom operations — because a workflow is a workflow: inputs, steps, decision points, outputs.

Lore — Institutional Memory

Decisions, historical traces, and prior art stored in a form both people and AI can query. How Foundry avoids repeating itself; how AI agents get context beyond their training data.

Forge — Visual Cockpit

The supervision layer. Forge renders Anvil workflows visually, surfaces AI actions in real time, and gives teams the control plane to supervise, approve, and intervene.

Temper — Measurement Loop

Continuous quality tracking for AI agents. Archives what agents do, grades outcomes over time, detects drift, and runs controlled experiments to find what improves it.

Why this matters

One platform.
Every product benefits.

Because all four products share the same foundation, improvements to any layer make every product better — automatically, without extra work.

01.Shared improvement cycles

A better memory model in Lore makes Forge answers sharper and Temper experiments more accurate. Improvements don't stay local.

See what this means
A better workflow primitive in Anvil makes every product that uses it faster. A tighter measurement methodology in Temper informs how Lore evaluates decision quality. No layer improves in isolation — every improvement fans out across the whole system automatically.

02.Cross-domain learning

Patterns that work in construction estimating inform financial review workflows. Every engagement raises the baseline.

How this compounds
Measurement techniques proven in one vertical generalize to others. A knowledge capture approach that works for a wealth manager's client data informs how we capture institutional memory for a general contractor. Each engagement compounds the platform's capability and yours.

03.No integration tax

The layers were designed to work together from the start. No glue code between memory, engine, cockpit, and measurement.

The technical detail
Each layer is thin and cohesive. Lore speaks a query protocol that Anvil, Forge, and Temper all understand natively. No adapters, no translation layer. The absence of integration tax means improvements to one layer appear in all the others without extra engineering work.

04.Customers inherit the platform

Your workflows run on this foundation. You don't operate the platform; you benefit from it.

What you actually get
When Foundry works with your business, the products we build for you compound with every engagement. The knowledge base grows. The measurement loop tightens. The workflow definitions get more precise. None of that requires you to manage the platform underneath — it compounds automatically.
Ready to go deeper?

See how Foundry deploys
this for you.

Customers benefit from this platform without understanding or operating it. Your tools run on the same foundation we use for everything else — and get better with every engagement.

See how an engagement works →