The Forge
Our production system
for shipping companies.
The Forge is how a small team ships like a large one: coding, creative, and growth agents, shared tooling, evals, and a weekly release cadence. Every Entatan project, partner or venture, runs on it.
agents in the fleet across code, creative, and growth
The build
Built for leverage, not headcount
The usual answer to more work is more people. We took the other path. From the first week, every repeatable task at Entatan got a spec, a tool, and an agent, and every agent got a test. The Forge is the accumulation of those decisions: the fleet, the templates, the evals, and the cadence that keeps it all honest.
It was not built as a product. It was built because a team of a handful of people had to run partner sprints and three ventures at once, and there was no other way to do it well. Everything in it earned its place by removing a bottleneck we actually hit.
Inside The Forge
Six systems, one pipeline
Each part is boring on its own. Together they let one team run engineering, marketing, and ventures without a handoff.
agent runs per week across partner and venture work
products shipped through The Forge in 2026
Coding, research, creative, and growth agents with scoped tools and budgets. Each has an owner, a spec, and a test suite.
Every prompt or agent change runs against recorded fixtures. A regression blocks the release, not the customer.
Systems
What runs on The Forge.
Agent fleet
Coding, research, creative, and growth agents with scoped tools and budgets. Each has an owner, a spec, and a test suite.
Evals before deploys
Every prompt or agent change runs against recorded fixtures. A regression blocks the release, not the customer.
Shared tooling
One CLI, one component library, one brand-token pipeline. New projects start from a working template, not a blank repo.
Model routing
Tasks route across model providers by quality, cost, and latency. We pay for what a job needs and cache the rest.
Weekly cadence
Monday plan, daily ships, Friday review. Partners get a changelog every week and a working build every day.
Observability everywhere
Every run is traced: tokens, cost, latency, outcome. Spend and quality are visible per project, per agent, per day.
How work flows
From brief to release
A project enters The Forge as a written brief with a measurable goal. Agents turn the brief into a plan, the plan into tickets, and tickets into pull requests, creative drafts, or campaign configs. People review at each gate. Nothing ships without a human owner signing off on it.
Releases go out daily to staging and weekly to production, with the same pipeline for a partner's app and for our own ventures. The changelog is generated from what merged, so what we tell partners is exactly what shipped.
People in the loop
Agents draft. People decide.
Agents are good at the first eighty percent: scaffolding, variants, research, boilerplate, the tenth ad concept, the migration nobody wants to write. They are unreliable at judgment. So we design the pipeline around review, not around trust. Every output lands in a queue with a diff, a cost, and a person whose name is on it.
That structure is also what makes the work auditable. When a partner asks what an agent touched, we can show the run, the inputs, and the reviewer. It is the same standard we would want from a vendor working on our own companies.
What is next
The roadmap
Three things are in progress. Longer-running agents that own a workflow for days rather than minutes, with budgets and checkpoints a person can inspect. Tighter evals for creative and growth work, where quality is harder to score than a test suite. And a partner view into The Forge so teams we work with can see runs, costs, and queues in the client portal.
We publish what changes in the changelog. If something here matters to your team, tell us and we will move it up.
Questions
Is The Forge a product I can buy?
Not as a standalone product. It is the system every Entatan engagement runs on, so you get it by working with us. Partners see its output in the client portal and in weekly changelogs, and keep everything it produces.
Which models does it use?
Several, chosen per task. Routing picks a model by quality, cost, and latency, and we swap providers when a better option appears. No partner is locked to one vendor, and we can pin a specific model on request.
Does my code run on your infrastructure?
Your code lives in your repositories and deploys to your cloud accounts. The Forge orchestrates work against them with scoped credentials you can revoke at any time. We do not keep a copy of your production data.
How do you keep agents from breaking things?
Scoped tools, spending limits, recorded evals on every change, and a human reviewer on every merge. Agents cannot deploy to production on their own. When something slips through, the trace tells us exactly what happened.
Can my team keep using The Forge after an engagement?
You keep every artifact: code, agents, prompts, evals, brand tokens, and campaign configs, all documented. Retainer partners keep running on the pipeline itself. We can also train your team to run the pieces without us.
Put The Forge to work
Bring a brief. We will scope a sprint, show you the pipeline, and ship the first release in two weeks.