Storage
NVMe
State and vectors
Agents accumulate context. Fast local storage, not a network volume billed per operation.
Agents need a machine that stays up, holds state and has an address of its own. A laptop that sleeps is not that machine. We are building one-click hosting for the agent platforms people already self-host, on the same unmetered servers as everything else we run.
Not yet available — you can run any of these on a standard VPS in the meantime
Two agent runtimes we build and operate ourselves. Both are aimed at long-running agents that need to survive a restart, keep credentials safely, and be reachable from outside.
Our own agent runtime, hosted and kept running for you rather than left on hardware you have to babysit. Survives a restart, holds its credentials, and is reachable from outside.
In developmentBuilt alongside OpenClaw for agents that run continuously rather than on demand, where the cost of stopping and starting is the thing you are trying to avoid.
In developmentAgents accumulate context. Fast local storage, not a network volume billed per operation.
Agents that crawl or call APIs all day do not generate a bandwidth bill here.
A server that does not sleep, close its lid, or lose its network when you leave the house.
Install anything. We do not inspect or restrict what your agents run.
These are established open-source projects with active communities. The plan is one-click installs on a server you control, with your data staying on it. Every one of them will also run on a standard VPS today if you would rather not wait.
All-in-one workspace for documents, chat and agents. Bring your own model or run one locally.
A full chat interface over local or remote models, with tools, retrieval and multi-user support.
Visual builder for LLM apps and agent workflows, with datasets and observability included.
Drag-and-drop builder for chains and agents, for people who would rather not write the glue.
Familiar chat interface supporting many providers at once, with search and saved presets.
General workflow automation with agent steps — useful when the agent is one part of a larger process.
Runs open models locally on the server, so nothing leaves the machine you rent.
Framework for multiple cooperating agents with defined roles and tasks.
One of the original autonomous agent runners, still widely self-hosted.
All open source, all self-hosted, all running on hardware you rent rather than a platform that reads your data
We would rather ship two of these properly than eight of them badly. If you have a preference, or something not on this list, say so and it will shape the order we build them in.