Help Center

A governed AI workforce for your business. Deploy a whole role-based AI team (a Team Pack) or individual agents — each hired for a specific job, working 24/7 under strict rules, with every action traced and risky steps sent to a human for approval. It runs on your own AI account, so you pay wholesale for AI, not marked-up retail.

Complete AI departments you deploy in one click. The SaaS Delivery Team is a software company in miniature — 18 roles from requirements and architecture to QA and security review. The Marketing Growth Teamis a revenue department — 20 roles from positioning and copywriting to outbound and analytics. Your core team comes pre-selected; uncheck anyone you don’t need, add optional specialists, deploy. Every member arrives with its job description, safety rules, and starter quality tests.

Yes. On the Agents page, click ⚡ Build from a website, paste a business’s website URL (and/or paste text directly), and the platform reads the site and builds a customized front-desk agent plus a ready-to-embed chat widgetin seconds — seeded with that business’s own name, phone, address, hours, and services. It only states facts it actually found on the page (it never invents a phone number or address), and you can open the result in the Studio to refine it. This is the recommended first step when standing up an agent for a new client or prospect.

An admin opens the Control Center, picks an agent, and defines its Spec: its identity, the deliverables it may produce, its rules, its negative space(what it must never do), and when it should escalate to a human. You then grow a small eval suite of real tasks and run it — an agent only earns more autonomy after it passes.

No. Setting up an agent is filling in plain-language fields (what it does, what it must never do). The platform compiles that into the agent's governed instructions for you.

Not without a human. Every external or consequential action goes through the Approval Gate — the agent drafts, a person approves. Agents earn the ability to act more independently only after proving themselves on evaluations.

It reasons only over knowledge you've given it (owned memory), scoped to its workstream. It cites its sources and, if it doesn't have enough context, it says so rather than guessing.

An eval is a graded test of a real task. Suites of them are your quality gate: an agent must pass before it's promoted to a higher autonomy level, and you re-run them after any change. This is how you know an agent is reliable before it touches customer work.

Yes — tap the 🎤 button next to any task box. In Chrome, Edge, or Safari it uses your browser’s built-in speech recognition, completely free. Prefer higher accuracy? Add a voice key (OpenAI or Groq) under Model Providers — it runs on your own account for pennies per minute. Your audio is transcribed and immediately discarded, never stored.

Yes — tap 📎 to attach up to 5 text documents (.txt, .md, .csv, .json) to a task: notes, policies, exports, transcripts. The agent reads the parts relevant to your task, and its claims can be tied back to exact quotes from your documents — receipts you can verify. PDF support is coming.

No. Isolation is enforced at the database level (row-level security) per tenant and per workstream — a user only ever sees their own lane's data. Workstreams cross only through a designed, approval-gated, audited handoff.

Several layers: inbound text is screened for prompt-injection, every model call is cost-capped and rate-limited with a per-day budget ceiling, poisoned knowledge is dropped before it reaches the model, and outbound content is checked for leaked secrets. Runs that hang are automatically swept and failed.

Every request carries a correlation ID that ties the action → the run → its steps → the logs, so an issue can be traced end to end in seconds. Failures are recorded per run, structured logs stream live, and error monitoring alerts the team.

Yes — an insert-only audit log and SOC-2 evidence ledger record every consequential action, approval, and policy decision. They cannot be edited or deleted from the app.

Solo $49/month ($39 annual) — 3 AI agents (from templates). Team $129/month ($99 annual) — Up to 15 AI agents. Agency $499/month ($399 annual) — Unlimited client workspaces. Every plan starts with a 14-day free trial, and we never charge per message. Details on the pricing page.

Cancel anytime in one click: Usage & Billing → Manage billing / Cancel— online, no phone call, no retention hoops. Cancel during the trial and you’re never charged. Double-charging is impossible by design: if your workspace already has a subscription, the system refuses to create a second one, and rapid double-clicks at checkout are caught at the payment layer.

Yes. Deploy a single template agent, or a Team Pack with only the members you check. Most teams start small, prove the ROI, then add more — your plan is the only limit, and safety features are included at every tier.

In an encrypted vault, write-only: after you save it, it is never displayed back to anyone — including you. Agents use it through a controlled path; it never appears in the app, in logs, or in your data export. Delete your workspace and the key is destroyed with it.

No. Your data is used only to answer your agents' tasks. The platform is LLM-agnostic and routes to approved providers under an allowlist.

Yes, anytime, no email required. Tenant Admin → Your data & privacy has two buttons: Download my data (everything we hold, as a file) and Delete forever(permanently erases your workspace, agents, runs, knowledge, members, and your encrypted AI key — and cancels your subscription first so you’re never billed after leaving). Full policy at seven16aiagent.com/privacy.