codehorse is now coderig. Same team and product, new name and address: coderig.io

Local coding agents on local hardware.

Stylized illustration of a small computer with the coderig logo.

coderig lets your company use powerful coding agents entirely locally. Affordable, with full data security.

Delivered as a preconfigured system with models and coding agents. Source code, prompts and results never leave your company network.

Book a demo

Preconfigured system

Hardware and software, ready to work together.

Local coding agents

Support for programming, refactoring and testing.

No cloud connection

All processing happens on your premises.

Coding agents for your team, including confidential projects.

Your most important projects can be the ones missing out on coding agents.

Coding agents could investigate bugs, implement changes and run tests. But confidential source code and internal policies often rule out external AI services. So your team keeps doing that work itself.

Cloud agents Processing at the provider
Your company network Cloud use not approved
Your teamConfidential source code
Agents remain out of reach

With coderig, agents can work on your confidential code too.

Models and agents run on local hardware inside your network, even without an internet connection. Your team delegates tasks and reviews the results. Code and prompts stay in-house.

Cloud agents No data exchange
Your company network
Code, prompts and results stay here.Works without an internet connection
Your teamConfidential source code
Task Result
coderig
Models & coding agents On your own hardware

You decide how your AI operates and how you use your computing capacity.

Your IT team chooses model versions, approves updates and can inspect local logs. There are no cloud fees per token, and no quota limits your usage.

Cloud agents No data exchange
Your company network
Your teamConfidential source code
Task Result
coderig
No cloud fees per token On your own hardware
Models
Your choice
Updates
Your approval
Logs
Stored locally

We engineer and optimise the entire system for local coding agents.

Operating system & driversInference engineModels & fine-tuningAgents & evaluationcoderig

Optimisation starts with the operating system.

We tune the operating system and drivers for the hardware being used. This creates the technical foundation for running models and coding agents locally.

Your IT team receives a platform that is already configured to work together.

The inference engine is tuned to the hardware.

We optimise the software that runs the models for the selected hardware. This work is part of the system we deliver, ready for your team to use.

Available computing power is put to work for local agents.

Fine-tuning adapts models to development tasks.

We fine-tune selected models and evaluate them against relevant software development tasks. Model selection becomes a carefully configured part of coderig.

Concrete tasks guide how we select and adapt models.

We evaluate the whole system on concrete tasks.

We benchmark models and agents as part of the complete system, assessing how they handle relevant development tasks on the selected hardware.

Evaluation covers the complete agent workflow.

Your team receives the result as a ready-to-use system.

Hardware, software and models arrive configured together. We keep developing the technical foundation while your team focuses on its development work.

New versions are available as tested updates. Your IT team decides when to adopt them.

Explore your configuration

The computing power your work calls for.

We configure coderig for your tasks, from a compact computer to a powerful rack. Your models and the way your team works determine the right setup.

An example of a compact coderig computer.

Compact enough for the office

For workloads that a single computer can comfortably handle.

Concept render of a larger rack configuration.

More capacity in a rack

For larger models and more agents working at the same time.

Bring local coding agents into your team’s everyday work.

We configure coderig around your work.

Together, we identify the tasks you want to delegate, how many developers will work at once and your IT requirements. We use this to find the right configuration.

Your team accesses the agents over your company network.

We deliver the prepared system. Connect coderig to power and your network; your developers open the local interface in their browser.

A described task becomes changes your team can review.

The agent examines the code, works through the task and runs tests. Your team reviews the changes and decides what to accept.

An office with three workstations and a small coderig on a sideboard. Each workstation uses agents running on this local device. Refactoring Tests Debugging
kundenportal Illustrative example · condensed Ready for review
You

Add a CSV import for customer records to our internal application.

Follow the existing architecture and permissions. Provide a preview, explain invalid rows and prevent duplicates, including on repeat imports. Implement the interface and backend, add unit and integration tests, and update the documentation.

Selected steps

Examines the data model and permissions

read_filesrc/customers/model.ts · src/auth/permissions.ts

Plans the work across components

update_planImport service, API, interface and tests

Builds the import, preview and validation

edit_fileapi/customers/import.ts · CustomerImport.svelte

Tests behaviour and component integration

run_testsunit/customer-import · integration/import-api

Prepares changes and documentation for review

review_changesdocs/customer-import.md

The implementation is ready for review.

CSV import ready for review✓ Tests passed
kunden.csvPreview with sample data
2 new 1 existing 1 invalid
  • Nordlicht SystemeK-104 · kontakt@nordlicht.example
    New
  • Westfeld TechnikK-101 · team@westfeld.example
    Existing
  • Kontur FertigungK-105 · kontakt@
    Check Email address is incomplete.
  • Hainwerk AnlagenK-106 · office@hainwerk.example
    New

Invalid rows are excluded. Existing customers are kept.

Agent and model running on coderigInternet traffic: 0 Bytes

What to know before getting started.

Does coderig need internet access?

No. The agent and models run locally, including in networks without internet access.

Which models can we use?

coderig runs open-weight models such as Qwen and GLM. Model and hardware choices depend on your workload.

What work does coderig take off your team’s hands?

We select and integrate the components, optimise the operating system, drivers and inference engine, and adapt and evaluate the models. Your team receives that engineering work as a configured system, without having to assemble the technical foundation itself.

How do we evaluate models and coding agents?

We benchmark models and agents against relevant development tasks. Evaluation includes their interaction with the inference engine and the selected hardware, testing the configuration that will be used in local operation.

How do you benefit from new models and optimisations?

We evaluate new model versions and continuously improve the software and models. Improvements are available as tested updates. Your IT team chooses which versions to adopt and when to install them through its own update process, including in offline environments.

How do updates reach an offline machine?

As signed bundles, transferred through your own update process. You decide when to install them.

How do we choose the hardware?

The models, repository sizes and number of concurrent tasks determine the configuration. We size it with you.

What does it cost?

Pricing depends on the hardware configuration. We provide a quote for your setup.

See what local AI can do for your team.

In a demo, we show coderig working through a concrete development example. Together, we work out which configuration fits your team.

Schedule a demo now

See coderig in action

Leave your contact details. We’ll get back to you personally.

Or email contact@coderig.io