Swisper Foundry

Foundry Manual

Ideas in. Production-ready code out.

Foundry runs a structured, configurable Product Development Lifecycle with a team of AI agents — business analysts, designers, architects, engineers in seniority tiers — steered by a handful of knowledgeable humans. The speed of agentic development, fused with a century of software-delivery discipline.

▶ The story in eight slides Start with the concepts See the architecture Create your first product

What makes Foundry different¶

Lifecycle A real lifecycle, not a prompt box

Products move through explicit phases — vision, design, planning, execution — each producing structured, schema-backed artifacts that downstream phases consume. Nothing is lost in a chat scroll.

Configurable The lifecycle is yours

The PDLC is data, not hardcode. Phases, artifact schemas, agent rosters and their connections are designed on a canvas in the Lifecycle Studio — tailored per organization and per product.

Agents A department, not a chatbot

Each agent has its own remit, prompt, model allocation and tool grants — engineers in seniority tiers from junior to expert. Humans review at the gates; agents do the work between them.

Execution Execution you can watch

Runs dispatch real engineering agents into isolated git worktrees. Live roster, streaming reasoning, costs and diffs as they happen — and runs are comparable, disposable entities.

Enterprise Credible by construction

Bring your own keys, or a vendor subscription — instance-wide or per user. Decisions and reviews on record, review gates only a human can pass, and a deployment story that fits inside your infrastructure.

The manual¶

Part I · Concepts What Foundry is

The mental model — products, lifecycles, phases, artifacts, agents, runs — and the architecture underneath it.

Part II · Admin Instance setup

Providers, models and pricing, the Copilot, MCP tool servers, users and roles — what an admin configures once.

Part III · Every user Your account

Bring your own key or subscription, how credentials resolve at runtime, and your GitHub identity.

Part IV · Products Creating a product

The onboarding saga — detect, confirm, materialize, verify — and the product settings that follow.

Part V · Lifecycle Studio The configurable PDLC

Design the lifecycle on a canvas, staff its agent roster, and shape prompts and artifact schemas.

Part VI · Daily work Running development

Mission Control, the design workbench, the execution cockpit, and the product knowledge registers.

Part VII · Operations Deploy and troubleshoot

The two-gate CI model, auto-deploy, migrations at boot, and where to look when something fails loudly.