Start with the questions below — they cover getting started, credits and billing, the Remote Gateway, security, and what to do when a task goes sideways. If your answer is not here, email support@frontbrain.ai and a real engineer will pick it up.
Our support requests are handled by real people. No frustrating AI nonsense here. We build an autonomous AI platform, and we still think a person should read your ticket — so when you write in you get an engineer, not a bot that summarises your question back at you and closes the thread.
No triage bot, no auto-generated first reply, no “was this helpful?” loop standing between you and an answer.
Tickets are answered by the people who build and operate FrontBrain AI, so a hard technical question gets a technical answer instead of a link to this page.
Every run leaves plans, tool calls, approvals, outputs, and verification results in the audit log. Send us the task and we can look at what actually happened.
If your ticket uncovers a platform bug, it goes to the people who can change the platform, and you get told when it ships.
Every plan includes email support and every ticket is answered by a person — there is no tier that cannot reach a human. See pricing for what each plan includes.
Private and self-hosted deployments can include custom support arrangements. Talk to us about what your environment needs.
Autonomous execution. Human support.
FrontBrain AI plans, executes, verifies, and documents real technical work. It runs SSH and infrastructure operations, code changes and deployments, database inspection, firewall and network review, cloud and virtualisation workflows, browser-based verification, and document generation — then proves the result before reporting it as done.
The product page lists the full capability matrix, and use cases shows what that looks like in practice.
Create an account at app.frontbrain.ai/signup and pick a plan. Hosted SaaS needs no infrastructure from you — you can start giving it work immediately.
Then connect the systems you want it to operate on. Connector credentials are added in Settings, stored in Vault, and scoped to your user and workspace. For systems that live on a private network, add a Remote Gateway.
Not for hosted SaaS. It runs in the browser and reaches anything that is publicly reachable — cloud APIs, SSH-reachable hosts, public endpoints.
You only install something when you want FrontBrain AI to operate inside a private network. That is the Remote Gateway: one small VM you control, one outbound connection, no inbound firewall changes. Details in deployment.
FrontBrain AI treats “done” as a claim that has to be proven. Before a task is reported complete it checks the result against the original request and collects evidence — files in place, services running, HTTP responses, screenshots, logs, command output.
If verification fails, the task is not silently closed: the failure is diagnosed, and for supported deployment workflows it can roll back to the previous known-good state.
Complex work can be split across specialist agents — infrastructure, coding, research, verification, security, and documents — running in parallel and then collected into one verified answer.
Multi-agent orchestration is a Business and Agency / MSP feature. Your plan sets how many agents can work at once.
No. Risk is classified by what an action actually does. Local artifacts pass freely, but deployments, DNS, firewall, email, billing, production installs, database writes, and power actions can require approval before they run.
On top of that: a remote write cannot execute until FrontBrain AI has inspected the same target in the same session, high-risk requests start in read-only planning mode, and approval can be required per call, per campaign, per workspace, or from multiple parties. The full control set is on the security page.
Credits are how usage is metered. They cover model usage, tool execution, and artifact generation — the real compute, third-party API calls, and execution resources a task consumes.
Each plan includes a monthly credit grant. Current allowances are on the pricing page.
Monthly plan credits replenish each billing cycle and do not roll over. Top-up credits you buy separately do not expire.
You can buy a top-up at any time, and you can opt in to soft overrun so a long-running task does not stall right at the end of a billing cycle. Overrun usage is real consumed service usage and may be recovered from future renewals, top-ups, or credit grants.
Yes, where supported. BYOK keeps model usage under your own provider terms, billing, and retention policy. Charges incurred through your own keys are billed by that provider directly and are not part of your FrontBrain AI invoice.
Payments are processed by Paddle, our authorised reseller and merchant of record. Paddle handles the transaction, tax, and invoicing, and Paddle receipts are the ones that appear on your card statement.
Plan changes and cancellation are handled from your account. Refunds follow the refund policy: requests must be made within fourteen calendar days of the transaction, and used credits, completed tasks, and consumed compute are not refundable.
If something was charged that you do not recognise, email support@frontbrain.ai with the date and amount and we will trace it.
The Remote Gateway is a small agent you run inside your own network. The FrontBrain AI control plane stays in the cloud; the gateway executes tools locally on your side of the firewall.
You need one when the systems you want operated are not reachable from the internet — internal servers, hypervisors, databases, firewalls, and client sites behind NAT. Without a gateway, FrontBrain AI can only reach what is publicly reachable.
No. The gateway makes a single outbound TLS WebSocket connection on TCP 443 and keeps it open. Nothing connects inbound to your network, so there is no port forward, no inbound rule, and no exposed listener to defend.
It also only runs the tools you whitelist for that site, and a gateway can be revoked from the dashboard at any time, which closes its session.
A small VM you control:
In the dashboard, go to Gateways and create a new one. Give it a name, and on plans that support it you can pin it to a specific client workspace. The dashboard then shows a one-line install command containing a single-use enrolment token.
Run that command as root on the host that will act as the gateway. The agent installs, verifies its release, registers the token, and opens its outbound session — at which point the gateway's status in the dashboard flips from pending to active.
The enrolment token is displayed once and is short-lived. If it expires or you lose it before running the command, rotate the token from that gateway's page and use the fresh command.
Work through these in order:
Still stuck? Email support@frontbrain.ai with the gateway name and the output of the install command.
Starter does not include a gateway. Pro includes one site. Business and Agency / MSP support multiple gateways — the current per-plan allowance is on the pricing page.
Pinning a gateway to a specific client workspace, which is what MSPs need to keep client sites separated, is available on Business and Agency / MSP.
Yes. Private and self-hosted deployment runs on customer-controlled infrastructure with your own Vault, Postgres, object store, and model routing, and can include a dedicated control plane, isolated toolrunners, private inference, and regional deployment. Talk to us about the requirements.
Connector credentials live in Vault and are scoped by user, workspace, and target. A tool that tries to use a connector the user does not own is rejected. Traffic is TLS in transit, and platform storage and artifacts use encrypted storage.
Yes. Conversation history, audit logs, artifacts, and connector access are tenant-scoped. MSP client workspaces go further, with separate connectors, users, permissions, and audit exports per workspace.
Every connector touch, tool invocation, approval, output, and verification result is written to an append-only audit log. Logs can be exported by tenant, workspace, user, connector, date range, or task, so a run can be reviewed after the fact.
Email security@frontbrain.ai — that inbox is for disclosure and goes straight to the platform team. For privacy and data-protection requests, use privacy@frontbrain.ai. Our privacy policy covers what we retain and why.
Usually because the platform stopped it on purpose. Repeated same-error loops, patterns that change no state, and approval-gate loops are detected and paused rather than burning time and credits on a task that is not progressing.
The other common reason is that the task hit an approval gate and is waiting for you. Check the run for a pending approval before assuming it failed.
That is a platform bug, and we want it. Send us the task so we can pull the plan, tool calls, and verification evidence for that run and find out where the verification gate let it through.
The more of this you can give us, the faster we get to the cause:
Please do not send passwords, API keys, or private keys in a support email. If we need credentials to reproduce something, we will tell you how to share them safely.
Four inboxes, so your message lands with the right people first time.
Product questions, technical help, gateway and connector trouble, billing queries, bug reports.
Vulnerability disclosure and anything security-sensitive.
Data protection, retention, and privacy requests.
Sales, private deployment, procurement, and everything else.
No deflection bot, no canned reply. Send us the task and we will look at what actually happened.
Less dumb. More done.