Understand the proposition.
What is Mentis8?
Mentis8 helps your firm become AI-native through a custom agent harness, inference capacity you own instead of per-token API billing, and control over enterprise data and execution.
Is Mentis8 available?
Mentis8 is pilot-ready. A pilot begins with workload and site qualification, a defined scope and agreed acceptance criteria. Contact Alex at alex@fireboard.ai to discuss fit.
How does Mentis8 make our firm AI-native?
We design a custom agent harness around your workflows, approved systems and knowledge. A pilot establishes the first useful outcome, its permissions and acceptance criteria before expansion across teams.
How does ownership change the economics?
Replace a provider’s charge on every local inference token with an investment in dedicated capacity. Within that capacity, more useful work does not create more API fees. Energy, software, service and maintenance remain costs; additional capacity requires investment.
Do we keep control of our data?
Yes. Mentis8 has no remote backend and no access to customer data. Inference, memory and learning run locally. Your chosen Slack, email and other connected services retain their own data paths; the runtime itself does not require a cloud service.
Is this just a server rack?
The hardware supplies capacity. The offering also centers on the operating layer, scoped enterprise connections and a qualified deployment. Specific delivery and service scope belong in the agreement.
Does local inference mean zero cost?
No. Local inference can avoid per-token API charges, but hardware, energy, software, maintenance and operations still have costs. Model licences and commercial terms also matter.
Can it go into any office?
No. Electrical capacity, heat rejection, access, floor loading and service space must be assessed. Modular installation is a design objective for qualified sites.
Which models can it run?
Models must be qualified against the selected hardware, runtime, licensing and workload requirements. The site does not promise a specific model version or measured throughput.
Can agents act across the entire company?
Enterprise-wide capability is a rollout goal, not automatic access. Each source, tool and workflow needs an explicit scope, owner and permission boundary.
Does on-premises mean fully air-gapped?
Not automatically. Inference location, network egress, updates and support access are separate decisions. An air-gapped requirement must be included in configuration and operational qualification.
What does the customer own?
The agreement should distinguish hardware ownership, enterprise data, software operating rights, third-party model licences and ongoing support. These are not interchangeable.
What does it cost?
Commercial pricing follows workload and site qualification. This site does not publish an unvalidated configuration price or treat an internal planning estimate as a quote.
What is the first step?
A private briefing around a specific workload, the people accountable for it, the data boundaries and the deployment environment.
Are these customer case studies?
No. Application examples illustrate workflows to evaluate. They do not represent named customers, deployed integrations or measured outcomes.
Do agents remember between tasks?
Yes. Agents retain useful knowledge and proven procedures, with provenance so their knowledge can be inspected.
How do agents know a task is complete?
They check work against the goal and observed environment, retaining verified progress and adapting when a step fails.
Does learning send our data to Mentis8?
No. Learning runs locally. Mentis8 has no remote backend and no access to customer data.