How Forgestus plans, builds, verifies and records every change — and the stacks and AI models it runs on. This page is for the engineers who will work with it.
No big-bang rewrite. Forgestus modernizes your monolith slice by slice — and proves each slice behaves identically before it ships. Every step durable, every retry idempotent, every cost in the ledger.
A read-only scan maps your architecture and business rules, ranks risk by churn × complexity × coverage, and produces a dependency-ordered strangler plan with difficulty scores. Nothing is written to your code.
Before touching anything, Forgestus runs your existing tests on the old .NET Framework runtime and pins the result — the behavior every migrated slice will be measured against.
Agents port the slice to modern .NET on a Windows runner — applying recipes like CoreWCF where they fit, and flagging anything non-portable for a human instead of guessing.
A green parity report — new runtime vs the pinned baseline — gates the merge. The cutover board moves the slice monolith → shadow → live. Repeat until the monolith is modern.
Every slice lands with a green parity report as its receipt — book a demo →
Node, modern .NET, classic .NET Framework on Windows Fargate, React, Next.js, Angular, Blazor, Flutter, Unity, WinForms/WPF, SSDT. The enterprise long tail, first-class.
Every transition durable in DynamoDB. Every retry idempotent. When an agent crashes, the orchestrator recovers without losing work. Crash-safe by construction.
Every model call, every retry, every dollar — written to AUDIT# rows that never expire. CSV export, regulator-ready (SIEM connectors on the roadmap).
Stuck epics auto-diagnose and fix themselves where safe. When an agent hits a target that doesn't exist, it pauses and asks instead of burning its budget — you /answer and it resumes. Every blocker surfaces a one-glance diagnosis with the exact fix command. No "wedged forever" mode, no log spelunking.
Add compliance: [hipaa] to forge.yaml — every agent inherits the coding controls. Enforced by the runtime, audited by the reviewers.
Anthropic, AWS Bedrock, OpenAI, or Vertex — your contract, your rate, your data-handling. We run the infrastructure; the flat platform fee covers it. Model tokens are the only usage cost, billed to you by your vendor — no bundled credits, no markups.
Classic .NET Framework 4.x on a Windows Fargate runner. Lambda dotnet10 on Linux. Flutter, Unity, Angular, Next.js — every stack with a first-class agent prompt and a deterministic QA runner.
Forgestus routes every agent to the model you choose — bring your own keys and switch anytime, per agent, per project. Run reviews on the strongest model you trust and bulk work on the fastest one you like; when a better model ships next month, it's a config change, not a re-procurement. Model tokens are billed to you by your vendor at your contract rate.
Anthropic keys on every plan; AWS Bedrock on Business+; OpenAI, Vertex, and Meta on Enterprise. Every model call — whoever serves it — lands in the same audit ledger with tokens and cost itemized.
The same deterministic agents that migrated your monolith deliver your everyday epics — across 11 stacks, with the same audit ledger and adversarial review gates. File a GitHub issue, type /proceed, and the pipeline ships it.
Open a GitHub issue. Describe the work in plain language. Forgestus's planner agent breaks it into stories, posts a plan, and waits for your /proceed.
Architect, backend, web, and mobile agents work on isolated story branches at the same time — each with full repo context, each running the right toolchain for the right stack.
Reviewer agents audit every PR on the strongest model you allow. halt verdicts trigger auto-retry of the implementing agent. advisory verdicts get logged. approve moves the epic forward.
Forgestus merges to develop, runs your deploy workflow, smoke-tests the deployed routes, and advances to the test environment on green. Failures route to devops-fix with the actual error in scope.
$4,000 a month puts a full delivery team of agents on your backlog, working in parallel across up to 15 repos, with every change reviewed and recorded in the audit ledger.
$4,000 across a 173-hour work month
75 epics included every month, each planned, built, reviewed and opened as a PR
With annual billing, 15% off
Plus model tokens, billed by your AI vendor at your own rate. Every plan runs on your own AI key.
Platform fee shown. Windows-runner epics count as two.
Book a free 30-minute demo. We’ll walk through a sample migration, then look at your system together.