Solutions
Private, safe AI for companies
Practical guides for companies that want AI agents to do real work without handing over their keys, their clients’ data or the final say. Written by the team behind MoltenRock.
- Guide · Private AIPrivate, safe AI for companies: a practical guideWhat “private AI” really means for a company: where the model runs, who holds the keys, who approves changes, and what gets logged. A practical setup for small teams on Mac.Read the guide →
- Guide · Shadow AIA simple shadow-AI policy for small businessesStaff are already using AI tools you never approved. A one-page shadow-AI policy for small businesses: approved tools, data rules, agents and approvals, and how to roll it out in a week.Read the guide →
- Guide · Law & accountingAI and client confidentiality: a practical guide for law and accounting firmsHow law firms, accountants and fiduciaries can use AI without breaching client confidentiality: where client data goes, which model to use, email risks, and a practical checklist.Read the guide →
- Guide · Human in the loopHuman in the loop, explained: when an AI agent should ask firstWhat human-in-the-loop AI means, which agent actions need a person’s approval, what a good approval looks like, and the patterns that fail. With a checklist.Read the guide →
- Playbook · MoneyAI agents and money: a refunds and disputes playbookHow to let an AI agent prepare refunds, dispute evidence and invoices without letting it move money on its own: rules, approvals, evidence, test mode and a checklist.Read the guide →
- Playbook · CloudflareLetting AI change your Cloudflare safelyA practical way to use AI agents with Cloudflare: read-only first, scoped tokens, a propose-review-approve-verify routine for DNS, SSL, firewall and Pages changes, and a checklist.Read the guide →
- Playbook · SecretsSecrets hygiene for AI agents: OpenClaw, Hermes and your API keysWhere AI agents leak API keys, and a routine to stop it: one vault, per-agent grants, logged use, rotation after exposure, verified cleanup and separate lanes for email and business tools.Read the guide →
- Playbook · EmailAI email triage without send accessLet AI read, sort, summarise and draft your email while sending stays human: a daily triage routine, why draft-only is the right default, choosing a cloud or local model, and a checklist.Read the guide →
- Playbook · PolicyAn AI agent policy you can build in five minutesA one-page policy for the AI agents that work in your company’s tools: which systems, which actions wait for a person, how keys are kept, what is logged. Build it, copy it, adapt it.Read the guide →
- Playbook · Buying10 questions before an AI tool touches your accountsBefore an AI tool or agent gets access to your payments, website, code or inbox: ten questions about keys, scope, approvals, logs and data, and how to score the answers.Read the guide →
- Playbook · IncidentsWhen an AI agent gets something wrongAn AI agent refunded the wrong order, deleted a record or emailed the wrong list. What to do in the first hour, step by step, and how to stop it happening again.Read the guide →
The products that do it
MoltenRock, MoltenRock Connect and MoltenMail put these guides into practice on a Mac: sealed keys, approvals before changes and a full log.
See the solutions on moltenrock.comRolling out to a team?
Email [email protected]. We reply within 24 hours and can help with setup, policy and custom integration.