System
Separate coordination, execution and durable state
The main server coordinates the programme. Remote sandboxes carry out builds, tests and review. Shared storage preserves the records needed to recover either layer.
A dedicated coordination layer
The lead agent, managers and coordination services run on the main server. They handle scheduling and programme decisions while builds and tests run remotely.
This separation keeps resource-intensive execution away from the services responsible for coordinating the fleet.
Remote execution capacity
Railway and Blaxel sandboxes prepare environments, index code, implement tasks and run verification. Capacity is provisioned against the schedule, with prepared spares available for replacement.
- Builds and tests never touch the main server
- Two providers, one common contract
- Prepared replacement capacity
State that survives a machine failure
PostgreSQL stores ownership and graph data, Git stores source, and artifact storage retains evidence. These records allow the fleet to recover coordination and continue from recorded checkpoints.
Hyperfide Open Source
The core workflow, open source.
Manage specifications, coordinate agents and verify results on your own infrastructure, using your Claude Code and Codex profiles.
Hyperfide Custom
Configured for your team.
Work with Sidiora Labs on deployment, execution capacity and review policies, with support for planning and running your first programme.