FAQ
Straight answers.
Everything here matches the services and pricing pages exactly. If a question is missing, ask it through the enquiry form and it gets a response within 24 hours.
Pricing
01What does an install cost?
A Solo Install is $3,000 fixed. A Team Install is $6,000 fixed. The add-on menu is fourteen items across four price lines: $400 for a standard connector, $700 for a complex build, $2,000 for a standalone security audit, and three items priced on application. Anything beyond the menu is also priced on application, anchored on $850 per day. Eleven of the fourteen add-ons carry a published price, so most builds can be totalled from the menu.
02What do retainers cost and what is in them?
Maintain is $350 a month: staging-tested updates and version pinning, with ticket support. Operate is $700 a month: everything in Maintain, plus monitoring and a monthly review, with minor config changes included. Partner is $2,000 a month: everything in Operate, plus two days of bespoke work each month and a priority queue. If you do not use both days, one carries into the following month and expires at the end of it. Nothing accumulates beyond that, because banked days we could not deliver in a burst would be a promise rather than a service.
03Who pays for the language model?
You do, directly. LLM provider costs and API keys stay in your account, on your own keys, and never pass through us. We cannot guarantee what a model says, and we say so up front.
04What will model usage cost us per month?
We do not publish a range, because it varies too much with what your agents do and how often they do it. Model usage is billed by your LLM provider to you directly, on your own keys, and never passes through us. Your provider's dashboard is the honest number.
Delivery
05How do I buy?
On the buy page, in four questions: who the agent is for, where it lives, what it should do first, and whether we keep it maintained. Each answer narrows the next, and the install, the framework, the integrations and the security posture are worked out from them rather than picked from a list. You see the total as you go, place the order, and the build starts from your answers.
06Do you host our agents?
No. Deployment is always on infrastructure you own: your cloud account or a VPS you purchase. Your credentials never leave your environment, during install or after handover.
07We do not have a platform engineer. Can we still run this?
Yes. The service is built for a team of roughly 10 to 200 people without a dedicated platform engineer. We deploy onto a VPS or cloud account you own, and we can walk you through setting one up.
08Which frameworks do you install?
Four, with more available on request including fully bespoke builds. QM, built at Y Combinator, is a multiplayer agent harness that runs in Slack and on the web, with scoped memory and permissions per person and per room; we install it where a whole team needs agents in one place. Buzz, by Block, is a workspace where agents hold their own identity and audit trail alongside people; we install it for teams that want agent work signed and traceable. OpenClaw, by the OpenClaw Foundation, is a personal agent runtime with a large skill ecosystem, for one operator working on their own workflows. Hermes, by Nous Research, is a personal agent that builds skills from its own use and reaches you through the chat tools you already have. All four are open source, QM, OpenClaw and Hermes under the MIT licence and Buzz under Apache 2.0; those upstream licences are where your rights to the framework itself come from. The frameworks are implementations, not the product: customisation is built at their extension points first, as skills, plugins and MCP servers. The service outlives any single framework, and framework migration is on the add-on menu.
09Where are you based?
Optaimi Ltd is based in the UK. The work is fully remote and we take on clients worldwide; wherever you are, deployment lands on infrastructure you own. An enquiry gets a response within 24 hours.
10If Optaimi disappeared tomorrow, would our deployment keep working?
Yes. Your agents run on infrastructure you own and your credentials never leave your environment. Most customisation is built at the framework's extension points, as skills, plugins and MCP servers. Where a requirement reaches the core, it is kept as a tracked patch against your pinned version and listed in the signed handover document, so another engineer can see exactly what differs from upstream and pick it up. Nothing about the deployment depends on us continuing to exist.
11Which chat tools do agents work in?
Team Installs surface in Slack. Microsoft Teams is not sold for a Team Install until we have verified it on the framework; an enquiry gets a written answer within 24 hours. A Solo Install can add a second chat surface as an add-on: Teams, WhatsApp, Telegram, Email or Signal. Those are examples of the class rather than the whole of it, so if the tool you use is not listed, name it in your enquiry and the price comes back with the response.
12Our system is not named on the add-on menu. Does that mean no?
Usually not. The systems named on the connector lines are examples of their class, not an exhaustive list, so if yours does the same job as the ones named, a chat surface, a CRM, accounting, an office suite or a helpdesk, it is normally on the same line and that is confirmed in the response to your enquiry. Two other cases sit on the menu without a fixed price: your own internal software with an API is reached through a bespoke MCP server, and a product that belongs to none of those classes is beyond the menu. Both are priced on application once scoped, anchored on $850 per day, because how long they take depends on the system rather than on us.
13Can we do a pilot before committing?
No. There is no pilot or trial tier. For anything above a Solo Install, a demo is available on request before you commit. A Solo Install is the smallest install we offer and ships at its published price.
14How long does an install take?
It depends on what your agents need to reach. An install can be handed over in as little as 24 hours; a wider scope takes longer, and you get the timeline for yours before you commit.
Safety
15What does hardened actually mean?
Least privilege by default: agents get access to the tools their job needs and nothing else, and actions with consequences wait for human sign-off. The posture you chose is recorded in a signed handover document. We refuse outright to run any deployment in a dangerous-equivalent posture, at any price.
16Can an agent send emails or spend money without approval?
Not under the default posture. Actions with consequences wait in a review queue for human sign-off: an agent can draft the email or stage the change, but a person approves it before it leaves. The signed posture document records what your deployment can and cannot do on its own.
17I have read alarming things about agent frameworks. Why put one in my business?
Those reports are the reason hardening is the product rather than a feature. An agent framework left on open defaults is dangerous, and we will not run one that way at any price. Every install ships hardened: agents get access to the tools their job needs and nothing else, and actions with consequences wait for human sign-off. The posture is recorded in a document you sign, and a dangerous-equivalent posture is refused outright.
18Can we see what an agent did, after the fact?
An agent activity audit trail is on the add-on menu as a complex build at $700 one-off: logging and retention of agent actions, set up on infrastructure you own. What can be recorded depends on what the deployed framework and the connected systems emit, and where a framework keeps its own trail, Buzz records agent identity and activity natively, the work is configuration and retention rather than construction. It is not part of an install or a retainer by default, and we never hold the log. The monitoring inside the Operate retainer is not this add-on and should not be read as a client-facing record.
19What happens if something goes wrong after handover?
The posture document says exactly what your deployment will and will not do, and you sign it at handover. After that point you are responsible for how agents are used; our obligation is hardening to the agreed posture plus training and guidance. Consequences of prompt injection or social engineering after sign-off rest with you, and training is part of every install so this stays unlikely.
20Who reviews an agent's permissions after handover?
You do. Every install ships with the hardened default in place, agents get access to the tools their job needs and nothing else, and that posture is recorded in the document you sign. After handover you can tighten or loosen those permissions, and revoking them is yours to do.
21Will this cost people their jobs?
Where these systems have been deployed, we have seen output and profitability go up, not headcount go down. That is what we have seen, not a promise. Training sessions are part of every install, and actions with consequences wait for a person's sign-off, so the work changes shape rather than disappearing.
Support
22What does support cover?
Every retainer includes 24/7 agent-assisted support with a guaranteed response within 24 hours. Human escalation is handled the next business day. Response is guaranteed, resolution is best-effort. In scope: any issue with the deployment as we installed and configured it, on the pinned stable version.
23What is not covered?
Client-initiated upgrades, third-party outages, LLM provider costs and model behaviour, updates you apply before our staging sign-off, data loss where the backup add-on was declined, consequences of prompt injection or social engineering after the posture was signed off, and new features or integrations, which are sold from the add-on menu rather than raised as support tickets. The full list is published because knowing the edges is part of what you are buying.
24What happens when the framework ships a breaking change?
It waits until we have tested it. Your deployment is pinned to a stable framework version, so an upstream release does not arrive on its own. Every update is tested on our staging before it is released to any customer and promoted only when needed, so a release that fails on staging is not promoted. Staging-tested updates are part of every retainer, starting with Maintain. If your deployment carries patches to the framework core, each patch is reapplied and staging-tested on each upstream release, which is why a patched deployment carries more upgrade work than an unpatched one. The mirror of this is in the published exclusions: an upstream release you apply yourself, before our staging sign-off, sits outside the SLA until it has been tested.
25Do you offer enterprise SLAs?
Not unless a specific deal justifies it. The standard SLA is published and applies to every retainer: response within 24 hours guaranteed, resolution best-effort, human escalation next business day.
26What happens after I send an enquiry?
You get a written response within 24 hours with a recommendation and the fixed price. If the work needs scoping, the response says exactly what happens next. If you go ahead, the agreement comes before any work starts.