OmniLink
Build an Agent
OmniLink
Free agents run while you’re connected. Paid agents run when you’re not. Every plan — including the free one — gives you unlimited agents, the full platform, and your own model keys. You pay your model provider directly; we never take a cut of your tokens.
Most platforms charge you per agent, or per minute of the agent talking. We charge for one thing: whether we will wake your agent on a schedule while you are not there. That is the work our infrastructure actually does, so it is the only thing the paid tiers gate.
The agent itself runs on your machine, through the OmniLink
connector. That is deliberate: it is why your tools, your files and
your robots never have to leave your network, and why a robot bridge
on 127.0.0.1 is reachable at all. Its memory lives with
us, so it does not forget between runs.
So be precise about what persistence means here. A scheduled agent survives closing this tab, closing OmniSim, and logging out. It does not survive the machine being switched off, or the connector not running — there is nothing on your side to execute the work. Install the connector as a background service and it starts again with your computer.
Everything else is uncapped on the free plan. Build fifty agents if you want. Wire them to OmniSim and rehearse a robot arm until it works. That costs you nothing, because it costs us almost nothing.
Build as many agents as you like and run them while you’re connected.
Three agents that keep running when you close the tab, on a schedule you set.
Fifteen agents we wake on a schedule while you are away.
Sixty scheduled agents, and a direct vote on what gets built next.
OmniLink is bring-your-own-key. Before your first agent can think, you connect a model provider — and we would rather you hear that here than discover it after signing up.
Two of the options cost nothing. A Google AI Studio key takes about sixty seconds and needs no billing card; OpenRouter connects in one click and has free open models. If this deployment runs Ollama locally, that is a third free path with no account at all. If you already pay for OpenAI, Anthropic, or xAI, paste that key instead.
We never mark up a token, at any volume, on any plan. Your provider bills you directly at their rates. We never see the invoice and we take no percentage of it. The subscription above is the only money that reaches us.
The usual way to price an agent platform is per minute of agent time. At the rates those platforms publish, Vapi charges $0.05 per minute of agent time even when you bring your own model key, and ElevenLabs charges $0.08 per minute. One agent answering for eight hours a day is roughly 14,400 minutes a month — about $720 at Vapi’s rate, and about $1,150 at ElevenLabs’.
Ours is $12 a month, flat, for three of them. We can do that because we are not reselling inference: your key, your quota, your bill. The subscription pays for keeping your agents alive when you are away, which is a much smaller number than the one per-minute pricing is quoting you.
Agent Builder, Agent Master and Agent Lord are the same plans at the same prices you signed up at. What changed is what they include: the old per-plan cap on how many agents you could create is gone on every tier, including Free, and the paid tiers now cover how many of your agents keep running once you disconnect.
There is nothing to do and nothing to migrate. Your plan, price and badge are untouched, and you gain the uncapped agent profiles.
OmniSim is an Apache-2.0 robotics simulator you run yourself. It is free, it is not metered, and nothing on this page limits it. Rehearse an agent against simulated hardware as much as you like — what the agent says it did gets checked against the simulated physics rather than taken on the agent's word. Rehearsal only: nothing built here has been run on a physical robot, and we do not claim sim-to-real transfer.
We do not currently meter simulated robot time, so we do not charge for it and it does not appear in any tier above.
Any agent configured to run without you present: a standing order on a schedule, an agent woken by a webhook, or one listening on a connected channel. An agent you talk to in the browser is not persistent, however long the conversation runs, and does not count against the limit.
No. Every plan, including Free, has unlimited agent profiles.
You are told which limit you hit and when it resets. Nothing is silently dropped and nothing is billed as overage — there is no usage-based charge on any tier, so going over costs you a wait, never money. The per-minute figures above are the ceilings we design and budget against; we would rather publish them than surprise you with one later.
Yes, from the dashboard, at any time. You drop to the free plan at the end of the period you have paid for and your agents stay where they are.