GRACE Developer
GRACE Developer, also called GRACE MLOps, is the MLOps module of 2021.AI's GRACE AI Platform. It lets data science and IT teams build, deploy and operate models on their own infrastructure, keeping their preferred IDE and Git.
What is GRACE Developer?
GRACE Developer — the vendor also calls it GRACE MLOps, and both names appear on its own page — is one of four modules of the GRACE AI Platform published by 2021.AI, a Copenhagen company that presents itself as a European AI governance specialist. The other three modules cover generative AI, AI security and governance; each can be integrated individually into an existing setup. This module is the development and operations half of the platform, aimed explicitly at data scientists and IT professionals who need to develop, deploy and operate AI at scale.
Its central promise is that the AI infrastructure comes out of the box. Rather than asking a team to assemble and maintain a stack, GRACE supplies one, so that data scientists spend their time on models instead of plumbing. Around that sit four claimed pillars: standardised components, replicability, usability, and freedom to keep your own IDE. Reproducibility is framed as an answer to staff turnover — standardised components let results be reproduced and consistency maintained across projects even when key people leave — and the same reasoning drives the onboarding argument, where documented workflows are meant to shorten ramp-up.
Practically, developers keep the editor they already use, version control runs on a full Git integration for change tracking, collaboration and rollback, and open-source compatibility is claimed as a design principle so teams can extend functionality without proprietary lock-in. Cost control is treated as a feature in its own right, with tooling to monitor and cap the spend of deployed models and applications and to allocate resources.
The module is deliberately infrastructure-agnostic. It runs on-premises, in a public cloud, in a private cloud, in hybrid form, and even in a fully isolated environment with no external network connection. It is also model-agnostic, integrating open-source and closed-source systems, with Mistral, Llama, Gemma, DeepSeek, OpenAI and AWS named on the vendor's own pages. At platform level it feeds an AI registry, role-based access control and cross-team workflows. Named customers include Maersk Tankers, SimCorp, Nuvve, Cobham Satcom, Prometric and Rigshospitalet. It is sold exclusively on quotation.
What it does
- Develop AI and machine learning models in your own choice of IDE
- Deploy models and AI applications into production on managed infrastructure
- Operate and monitor deployed models continuously
- Version and collaborate on code through full Git integration
- Build AI applications with the bundled Application Builder
- Track and control the cost and resource consumption of deployed models
- Run the whole stack on-premises, in a private cloud, in a public cloud or fully isolated
When to use GRACE Developer / When not to
A quick filter to help you decide if GRACE Developer is the right fit.
When to use GRACE Developer
- Enterprise data science teams that need to move models out of notebooks and into supervised production
- Organisations in regulated sectors — finance, healthcare, public administration, legal, manufacturing, transport — where every model has to be registered and defensible
- Companies with strict data sovereignty requirements, since the module runs on-premises, in a private cloud or in a fully isolated environment
- Engineering teams that refuse tooling lock-in and want to keep their own IDE, their Git history and open-source components
- Platform and MLOps engineers asked to supply ready-made AI infrastructure instead of assembling and maintaining it themselves
When not to use GRACE Developer
- Individuals and small teams: there is no self-service sign-up, no published price and no way to start without talking to sales
- Anyone wanting to evaluate the product alone first, since neither a free plan nor a free trial is advertised
- Developers looking for a documented public API to integrate against, as no API documentation page exists anywhere on the site
- Buyers who need contractual terms up front, because the site publishes no terms and conditions and no legal notice at all
- Teams needing a mobile or browser-based companion app, none of which the vendor offers
How to use GRACE Developer
A typical end-to-end flow, from setup to results.
- Identify which of the four GRACE modules you need, since Developer can be bought on its own or as part of the full platform
- Contact the vendor through the quote request form, the contact page or the meeting booking form — there is no online sign-up
- Discuss requirements with the sales and technical team during an introductory meeting
- Agree a bespoke commercial proposal, as no price list is published
- Choose the deployment model: on-premises, public cloud, private cloud, hybrid or fully isolated
- Have the module deployed into the chosen infrastructure, or integrated into an existing solution
- Connect your existing Git repositories and let developers keep their preferred IDE
- Set up role-based access, workflows and registration of models in the platform's AI registry
- Develop, test and deploy models and applications from the workbench and the Application Builder
- Monitor deployed models, resource allocation and running costs on an ongoing basis
Pros & Cons
Pros
- Genuine data sovereignty: on-premises, private cloud and fully isolated deployments are core to the offer, not an afterthought
- No proprietary lock-in claimed — open-source compatibility by design, your own IDE, your own Git
- Model-agnostic, supporting both open-source and closed-source systems including Mistral, Llama, Gemma, DeepSeek and OpenAI
- Governance is built in rather than added later, since models developed here feed the platform's AI registry
- Infrastructure supplied ready to use, reducing the maintenance burden on the team
- Cost monitoring and resource allocation treated as first-class features
- Established European vendor with named, checkable customer references and published client stories
Cons
- No public pricing whatsoever — no figure, no range, no order of magnitude
- Neither a free trial nor a free plan is advertised, so the product cannot be evaluated independently
- No public API documentation exists, even though the vendor sells API access
- The site publishes no terms and conditions and no legal notice at all, which is unusual for an enterprise purchase
- No data processing agreement is published or offered, and no named list of subprocessors exists
- The vendor takes no written position on whether customer data may be used to train models
- The site contradicts itself on its own contact details, listing several phone numbers and two spellings of its address
Pricing & Plans
No price is published. GRACE Developer is sold on quotation only: the pricing page invites visitors to request a custom pricing plan, and every module, as well as the full-stack platform, carries a Request a quote button rather than a figure. Neither a permanent free plan nor a time-limited free trial is advertised. Prospective buyers should therefore expect a commercial discussion before obtaining any cost indication, and no starting price, currency or billing unit can be stated here.
Data, GDPR & hosting
A consolidated view of how GRACE Developer handles your data.
GDPR overview
GDPR coverage is explicit and unusually detailed. An entire section of the privacy policy is devoted to it, and a Data Protection Officer plus a dedicated privacy team are appointed, reachable at dpo@2021.ai with a postal address and phone number in Copenhagen. The policy describes a company-wide information audit, retention and erasure procedures, breach handling, free subject access requests answered within thirty days, Article 35 impact assessments for high-risk processing, Article 9(2) handling of special category data, and safeguards for transfers outside the EU through binding corporate rules or standard clauses. No Article 27 representative is named, which is consistent: the vendor is established in Denmark. One caveat: the document still refers to the May 2018 GDPR deadline in the future tense.
Who owns the data?
Under the privacy policy in force since 20 November 2023, the customer remains the data controller and 2021.AI positions itself strictly as a processor, stating plainly that it is not a data controller of any Content. The vendor undertakes not to view, access or use customer Content except where genuinely required to deliver the product or to handle a support request, and treats a support request as express permission to access Content for that limited purpose. It also commits to neither give, sell, share nor trade Content outside the cases the policy lists. Where the module is deployed on-premises, the material stays inside the customer's own infrastructure.
Reuse rights
The policy frames use narrowly: personal information serves the service or transaction it was provided for, and the vendor states it will not share, sell or rent it. Declared purposes cover handling enquiries and orders, product support and updates, marketing, billing and collection. Disclosure is possible to payment, hosting and archiving providers, to reselling partners, in a merger, acquisition or insolvency, and where the law compels it. Customers remain free to reuse their own content, since the vendor claims no ownership over it. One silence deserves attention: nowhere in the public documentation does 2021.AI state whether customer data may be used to train AI models, and no exclusion mechanism is described. Absence of a claim is not a commitment either way.
Data retention & training
Hosting summary
Two distinct layers should not be confused. For the website itself, the privacy policy states that personal information collected is stored and processed on servers located primarily in Denmark, and in other countries covered by the General Data Protection Regulation; visitors residing outside Denmark are deemed to consent to that transfer. Transport is protected by SSL, and stored content and personal information by 256-bit AES encryption. For the product, the picture is different and largely in the customer's hands: GRACE Developer is infrastructure-agnostic and can be deployed on-premises, in a private cloud, in a public cloud, in hybrid form or in a fully isolated environment with no external network connection. In an on-premises deployment the data never leaves the customer's own infrastructure, and the vendor states as a selling point that no data transits to external model providers or third-party AI services. Jurisdiction for the product therefore follows the deployment choice rather than the vendor's location.
Things to keep in mind
Risks and trade-offs to weigh before adopting GRACE Developer.
- No price is published anywhere, making budgeting impossible without a sales conversation
- The pricing page's source code contains leftover template plans with figures that do not describe this product at all; anyone finding them would be seriously misled
- No terms and conditions and no legal notice exist on the entire site, so the contractual framework is invisible before negotiation
- The regulatory frameworks displayed across the site are what the software helps you comply with, not certifications held by the vendor — an easy confusion when the product sells compliance
- No published data processing agreement and no named subprocessor list, leaving a gap to close contractually
- The vendor's silence on training models with customer data leaves a decision to the contract rather than the policy
- The privacy policy has not been substantively refreshed since November 2023 and still refers to the May 2018 GDPR deadline in the future tense
Setup & Integrations
Technical difficulty
Not a self-service installation. Deployment is handled by the vendor after contracting, so the practical difficulty lies less in installation than in the enterprise project around it: choosing between on-premises, public cloud, private cloud, hybrid or isolated infrastructure, then connecting existing Git repositories and setting up access roles. The vendor argues that supplying infrastructure out of the box and providing documented, standardised workflows shortens both setup and onboarding. Expect to involve data scientists and IT professionals, and to treat it as a project rather than a sign-up.
Deployment
Integrations
Behind GRACE Developer
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is GRACE Developer?
Who is it aimed at?
How much does it cost?
Is there a free trial or a free plan?
Can it run on our own infrastructure?
Which AI models can be used with it?
Can developers keep their existing tools?
Is there an API?
Where is data hosted?
Does the vendor train models on customer data?
Should you pick GRACE Developer?
GRACE Developer is a coherent and credible piece of enterprise tooling. Its functional story holds together: ready-made AI infrastructure, standardised and replicable components, freedom to keep your own IDE and Git history, open-source compatibility, multi-model support, and cost controls that are treated as a feature rather than an afterthought. The sovereignty argument runs through the whole offer without wavering — on-premises, private cloud and even fully isolated deployments are genuinely available, which matters a great deal to the regulated European organisations the vendor courts. Its customer references are named and checkable.
The reservations are commercial and contractual rather than technical. Nothing about the price is public: no figure, no range, no trial, no free tier, and no way to try the product without entering a sales cycle. More surprising for an enterprise purchase, the site publishes no terms and conditions and no legal notice at all, no data processing agreement is offered, and no list of subprocessors is named. The vendor sells API access through its gateway yet documents no API publicly. And on the one question every buyer of an AI platform now asks — whether customer data may be used to train models — the public documentation is simply silent, neither claiming the right nor renouncing it.
One point deserves care when reading the vendor's own material: the regulatory frameworks displayed throughout the site are what the software helps you comply with, not certifications the company holds. That distinction is easy to blur when the product being sold is compliance itself.
The verdict follows naturally. If your organisation has an enterprise procurement process, real sovereignty constraints and a data science team ready to industrialise, this is a serious candidate. If you want to assess a tool by yourself before speaking to anyone, look elsewhere.
- Choosing a selection results in a full page refresh.
- Opens in a new window.