CZAR FORGE
Coding agents on your own repositories, models tuned and evaluated on your own code and data, and the agents around them, on one sealed appliance inside your perimeter. For technology and AI teams only.
- WHAT IT IS
- Czar, configured for technology and AI teams
- FOR
- Software, platform, data, and machine-learning teams
- RUNS
- Coding agents, the repository index, tuning, evaluation, serving
- BOUNDARY
- Agents propose; a named engineer merges or releases
- POSTURE
- No route out, or only the systems the team declares
- BOUGHT
- One subscription for one configured appliance
WHAT A TEAM GETS
The whole build loop, on one node the team holds.
Code in, a tuned model, a scored result, an agent that proposes the change. None of it leaves the building to happen.
Coding agents on your own repositories
Agents read, write, and test code against repositories that never leave the building. They open the change; they do not merge it.
The repository index
Code, design notes, tickets, and runbooks questioned in place, with the passage the answer came from.
Models tuned on your code and data
Open-weight models adapted to the team's own repositories and conventions, on the appliance. The adapter stays inside.
Evaluation before anything is promoted
A harness on the appliance scores models, prompts, and agents against the team's own test sets and past resolved work.
Serving for your own applications
Internal tools and products call the models on the Forge over the team's network, with no vendor in the path.
The team's agents in Agency Studio
Review, build-and-test triage, incident diagnostics, release notes: bounded, approved, versioned, and replayable.
WHO IT IS FOR
Built for the teams that build.
Czar Forge is Czar configured for the people who write software and build models: engineering, platform, data, and machine-learning teams. Every other kind of team gets the same appliance configured for its own work, as Czar.
- WHO
- Technology and AI teams only: software engineering, platform, data, and machine-learning teams. Every other team runs on Czar.
- WHAT IT IS
- Czar, configured for that kind of team. The same sealed appliance, loaded for code, models, and the agents around them.
- HOW IT IS BOUGHT
- The same subscription as any Czar, for one configured appliance: hardware, platform license, and sustainment in one line, fixed for the term.
- LATER
- When the team outgrows one Forge, more Czars join it as a Grid. Nothing is rewritten.
THE BOUNDARY
An agent proposes. A named engineer merges.
Agents on a Czar prepare and propose; a named person approves. Nothing client-facing, customer-facing or production-changing leaves the appliance without that approval, and every step of it is on a record the team holds.
PROPOSE
An agent opens the change
The diff, the tests, the run, and the evidence, assembled on the appliance and held in a queue.
APPROVE
A named engineer merges
Merge, promote, or release is a person's action, written beside the change with the approver's name.
ROLL BACK
Every version is kept
Agents, prompts, adapters, and the record are versioned. Any run can be replayed; any version can be restored.
INSIDE THE PERIMETER
The source, the models, and the record stay where the team is.
SOVEREIGN
No route out
The Forge runs with no connection at all. Updates arrive as signed bundles the team carries in.
DECLARED
Only the systems the team names
Source control, build and test, ticketing, the log and artifact stores: declared one by one at commissioning. Local Bridge denies the rest.
KEPT
The material stays
Source, datasets, adapters, evaluations, and the record of every run stay on the appliance the team holds.
READING
Czar Forge in the library.
THE SAME APPLIANCE
Czar, Czar Grid, Czar Keep, Czar Chassis.
Czar Forge is Czar for one kind of team. Czar Grid is several Czars joined into one pool for a facility. Czar Keep is that Grid in one chassis, for the data center. Czar Chassis is the Grid shipped by a vendor under its own name. All of them are this appliance.
- Substrate is what the Forge runs: Inference, Retrieval, Memory, Adaptation, Evaluation, Watchman, and Local Bridge.
- Agency Studio is where the team authors, bounds, approves, and records its agents.
- E31 Consulting builds the first agents and the evaluation set with your team, on your Forge.