The facts, with their sources

What it does

Done-for-you service: YesThis is a services business: Markovate's own engineers run discovery, architecture, build, integration and production deployment, and can be embedded onsite or remotely rather than handing you software to operate.Markovate's forward deployed engineers work alongside your teams, onsite or remotely, to turn complex business challenges into production-ready AI…markovate.com/forward-deployed-ai-engin…

How much it does unattended

Runs autonomously: PartlyThe agents they build are described as acting autonomously on multi-step workflows, but Markovate also states it designs human-in-the-loop checkpoints so clients keep approvals and exceptions, so autonomy is scoped per engagement.We design agents with human-in-the-loop checkpoints, so your team stays in control of approvals, exceptions, and sensitive actions.markovate.com/ai-agents-development-com…
Multi-agent: YesMarkovate says it engineers multi-agent architectures and orchestrates interdependent tasks across departments, and lists multi-agent systems among what its forward deployed engineers build.We design and engineer intelligent AI agents and multi-agent architectures capable of reasoning, planning, and autonomous action.markovate.com/ai-agents-development-com…
Agent permissions: PartlyGuardrails, access controls and execution controls are described as part of the architecture stage, but the site names no permission model a buyer can inspect before an engagement defines it.We define agent roles, decision logic, guardrails, and governance that meet enterprise compliance.markovate.com/ai-agents-development-com…

Which models it runs on

Claude: YesAnthropic's Claude models are named in Markovate's own model inventory and in its statement that it fine-tunes and integrates leading LLMs for domain-specific client needs.We fine-tune and integrate leading LLMs like GPT, Claude, and Llama for domain-specific needs, or develop fully custom Generative AI systems when…markovate.com/generative-ai-development…
GPT: YesOpenAI GPT models are listed in the AI-MODELS inventory and in the same sentence about integrating leading LLMs, and a separate services page describes integrating GPT into client products.Our team builds and fine-tunes LLMs, integrates GPT and other foundation models into your products, and develops adaptive systems that keep learning…markovate.com/services
Open models: YesOpen-weight families appear in the model inventory, with Meta's LLaMA and Mistral both named, and Mistral explicitly described as an open model they work with.Mistral 7B A lightweight, high-performance open model optimized for efficiency and fast text generation.markovate.com/generative-ai-development…
Model choice: YesMarkovate states it chooses models, frameworks and infrastructure from enterprise requirements instead of standardising clients on one AI ecosystem, which puts model choice inside the engagement.We select models, frameworks, and infrastructure based on enterprise requirements rather than forcing every customer into a single AI ecosystem.markovate.com/forward-deployed-ai-engin…

Where you use it

In a browser: YesMarkovate's own platforms run as hosted web applications where teams upload complete drawing sets, and the services page states they are deployed in Markovate's cloud, the customer tenant or on-prem.Proven AI platforms for engineering, construction, and manufacturing teams, deployed in our cloud, your tenant, or on-prem.markovate.com/services

Whose machine it runs on

Self-hosted: YesBoth the platform page and the engineering page offer customer-infrastructure deployment, including private cloud, on-premise and air-gapped environments, though the site qualifies it by technical feasibility.Depending on technical feasibility, we support cloud, private cloud, hybrid, and on-premise AI deployment architectures.markovate.com/forward-deployed-ai-engin…

What it costs to run

Free tier: NoThere is no self-serve or free entry point: every call to action is a demo request or strategy call, and the lowest-commitment route Markovate publishes is a scoped four to six week paid pilot.See real impact on your workflows before full commitment — through a scoped, low-risk pilot built on your actual data.markovate.com
API access: PartlyThe AI Blueprint Classifier platform documents REST APIs, webhooks and structured outputs for downstream systems, but no public API exists for the consulting side and no developer documentation is published.The platform supports JSON, CSV, and Excel outputs, REST APIs, webhooks, and integrations with systems such as Windchill, Teamcenter, Procore,…markovate.com/ai-blueprint-classifier

Buying it for a team

A company can buy it: YesThe entire offering is sold to organisations through enterprise engagements, with engagement models for onsite, hybrid and dedicated remote teams; there is no individual purchase path.Markovate offers forward deployed engineering services across the enterprise AI implementation lifecycle, from custom AI development and AI agent…markovate.com/forward-deployed-ai-engin…
Admin controls: PartlyAccess controls, governance and execution controls are designed into each build and the platforms use tenant isolation, but no customer-facing admin console or policy settings are documented.We apply Responsible AI frameworks, implement access controls, and ensure full transparency and auditability throughout each agent's lifecycle.markovate.com/ai-agents-development-com…
Audit log: PartlyThe drawing platform keeps results traceable to the source drawing and advertises a full audit trail, and agent builds claim lifecycle auditability, but no administrative action log is described.Auditability & Traceability Keep results connected to their source drawing and location so teams can review how information was derived.markovate.com/ai-blueprint-classifier

What happens to your code

Opt out of training: YesThe platform FAQ states plainly that customer drawings are not used for shared or cross-customer model training, so client data does not feed another customer's model by default.Is our drawing data used to train models for other customers? No. Customer drawings are not used for shared or cross-customer model training.markovate.com/ai-blueprint-classifier
Data residency: PartlyData-residency and deployment options are offered for enterprise requirements, including on-premise and air-gapped builds, but no country, region or cloud jurisdiction is named on the site.The platform uses tenant isolation, with deployment and data-residency options available for enterprise requirements.markovate.com/ai-blueprint-classifier
Getting out: PartlyPlatform output can be exported to Excel, CSV or JSON and pushed to downstream systems, and engagements end with documentation and knowledge transfer, but no contractual data-export or termination terms are published.Export takeoff quantities to Excel or CSV, or connect them to estimating tools through integrations, so they carry straight into pricing and bidding.markovate.com/ai-blueprint-classifier
Certifications: YesMarkovate holds ISO 9001:2015 and ISO/IEC 27001:2022 certification and states its practices align with GDPR and HIPAA; no SOC 2 report is claimed anywhere on the site.We are ISO 9001:2015 certified for global quality standards and ISO/IEC 27001:2022 certified for strong information security management. Our…markovate.com/ai-agents-development-com…