Everything you need to run a browser fleet
Machineuse gives you the parts that are hard to build well — scheduling, isolation, a snapshot lifecycle, healing, and brokerless messaging — as one self-hosted, MIT-licensed system.
Machineuse's core features are a distributed control plane, load-aware scheduling, a brokerless NNG mesh, systemd-nspawn isolation,snapshot-based dormancy, real-time metrics, auto-healing, and MCP tooling — all self-hosted with no SaaS dependency.
Scheduling & scale
Place work where there is real capacity, and add capacity by adding machines.
Distributed control plane
A central control plane tracks every worker and instance and decides where each browser runs. Add worker nodes and scheduling adapts automatically — no manual sharding.
Load-aware scheduling
Placement is by node capability and live utilization, not round-robin. Sessions land where there is genuine capacity, keeping latency predictable as the fleet grows.
Brokerless NNG mesh
Nodes coordinate over a brokerless NNG (nanomsg-next-generation) mesh. No RabbitMQ, Kafka, or Redis to deploy, secure, and babysit as a single point of failure.
Isolation & lifecycle
Every browser is contained, and idle ones stop costing you resources.
systemd-nspawn isolation
Each browser runs in its own systemd-nspawn container with a dedicated filesystem and resource envelope — clean, reproducible, and isolated from every other session on the node.
Snapshot-based dormancy
Idle instances are snapshotted to dormancy and their live CPU/memory reclaimed. On the next request, the snapshot is revived on demand — no cold rebuild, and far more logical sessions than you can keep live.
Per-session containment
Cookies, storage, and untrusted pages stay inside a session. One browser cannot see or interfere with another sharing the node.
Operations & integration
Observe the fleet, recover from failures, and drive it programmatically.
Real-time metrics
Per-node time-series metrics stream to the control plane, so it always has a current view of utilization to schedule against.
Auto-healing
Failure detection triggers automatic recovery, including migrating instances off a dead worker so the fleet keeps running without manual intervention.
MCP tooling
The project includes MCP-oriented tooling so agent frameworks and computer-use workloads can request, drive, and release isolated browsers programmatically.
Machineuse orchestrates the fleet — scheduling, isolation, and lifecycle. You drive the browsers with your own automation tooling; see the docs for supported drivers and connection details.
Get startedSee the architecture →
Related: use cases · comparison · FAQ