#1 deploy
Deploys in one click
You pick a name and a model, the agent comes up on its own. No servers, no YAML, no wrestling with systemd.
Cloud · in development
CapyAgent
lives in the cloud
An agent that does not switch off with your laptop. It runs around the clock, remembers everything between sessions and answers in whichever channel you happen to be in.
Plainly: the cloud does not exist yet. What follows describes what we are building, not what already works. The form is a place in the queue for early access, not a purchase.
- Self-hosted install works today
- Cloud: in development
- No billing and no account area
Capabilities
What the cloud will give
The list of things this is being built for. Not one item has shipped yet; the status of each will show up in the changelog.
#2 always
Runs while you sleep
The schedule executes on the cloud side. You can close the laptop and the morning summary still arrives.
#3 memory
Memory lives with the agent
Not with the device. Come in from your phone, the browser or Telegram - the conversation continues from the same place.
#4 channels
Every surface at once
Telegram, Discord, the Bitrix chat, a web page. One agent, one memory, a separate history per channel.
#5 secrets
Keys are stored apart
Credentials for your systems are held encrypted and handed to a connector only for the duration of a call. Each gets its own and sees no one else's.
#6 isolation
A container per account
Neighbours do not share your process. There is no multi-tenancy in the core yet, and it is the first thing that has to be finished for the cloud.
Status
What stands in the way right now
Keeping something that does not exist in a description is easy right up until the first client. So the list of open holes sits on the public page rather than in private correspondence.
- There is no multi-tenancy in the core. Contrary to an old line in the README, this version targets a personal or dedicated perimeter. For now each client is a separate process.
- There is no billing. No charges, no account area, no usage counter. Before a cloud that has to be built end to end.
- Container isolation has not shipped. Proving that one agent cannot reach another one's data is separate work with a separate risk.
- Where the machines physically sit is decided together with the first client: 1C data may not be held on our server, and that constraint is not going anywhere.
While there is no cloud, a working path is available: the agent is installed on your side as a single binary. How that is done is on the install page, and timelines and prices are in pricing.
Questions
About the cloud
When does it launch?
We name no dates until multi-tenancy and isolation are closed. People on the waiting list hear first, once there is something worth showing.
Can I use it now?
Yes, self-hosted. The core already runs as a service on a production server and executes scheduled jobs. The difference is that the server is yours, not ours.
Will the data sit on your side?
In the cloud version yes, and that is a deliberate trade. If the trade does not suit you, the self-hosted path remains: there nothing leaves at all.
What will it cost?
Undecided. Rollout and support prices are published in pricing; the cloud price will be worked out separately once the cost of a machine is clear.
How is it different from installing on my own server?
Not at all in what the agent can do, and entirely in operations. Your own server means your control and your support burden. The cloud removes the second half and adds the question of trusting us.
Early access
Join the waiting list
Write which systems you have connected and what the agent should be doing without you. Those answers decide what we build first.