x86_64
Linux AMD64
For 64-bit Intel and AMD machines, including WSL2 on x86_64.
Download AMD64 0.4.0-alpha · .tar.gz archiveNucleus 0.4.0-alpha
Your hardware. Your cloud.
Give these instructions to your coding agent to install the current Nucleus alpha.
Alpha 0.4 adds an explicitly configured CPU model worker for approved text models. Positronic, powered by Nucleus, provides model controls, shared organization documents, and scoped API and MCP access. Enrollment, resource telemetry, and node pairing remain available.
These packages run on Linux and WSL2. Go and Python are not required to run the included binary.
Run uname -m in your Linux or WSL2 terminal to check.
x86_64
For 64-bit Intel and AMD machines, including WSL2 on x86_64.
Download AMD64 0.4.0-alpha · .tar.gz archiveaarch64 / arm64
For 64-bit ARM Linux machines.
Download ARM64 0.4.0-alpha · .tar.gz archiveSave your archive and SHA256SUMS in the same directory. The first command below checks the archive. Continue when it reports OK.
Run the matching commands from the directory containing your downloads. On Windows, use a WSL2 terminal.
sha256sum --check --ignore-missing SHA256SUMS
tar -xzf nucleus-0.4.0-alpha-linux-amd64.tar.gz
./nucleus-0.4.0-alpha-linux-amd64/bin/nucleus init --name workstation
./nucleus-0.4.0-alpha-linux-amd64/bin/nucleus help
sha256sum --check --ignore-missing SHA256SUMS
tar -xzf nucleus-0.4.0-alpha-linux-arm64.tar.gz
./nucleus-0.4.0-alpha-linux-arm64/bin/nucleus init --name workstation
./nucleus-0.4.0-alpha-linux-arm64/bin/nucleus help
Initialization creates a persistent identity for this installation. It does not join another machine automatically.
Positronic is the paid managed service powered by Nucleus. For optional managed enrollment, use the installed CLI from a shell where nucleus is on your PATH.
Manage your organization at positronic.app/dashboard. The CLI address below remains the supported enrollment endpoint.
nucleus cloud connect --url https://nucleus.parallax.link/dashboardOpen the approval link, sign in with Nexus, and choose your organization. An owner or administrator must compare the machine fingerprint shown in the browser with the CLI before approving.
nucleus cloud runAfter approval, keep this command running in the foreground. It does not install a background service.
For connection details and an optional systemd service, read the hosted installation guide.
Model serving requires a locally configured CPU runtime and approval for the catalog’s exact model artifact. Enrollment and telemetry alone do not start model serving.
Give the versioned agent instructions to your coding agent, or follow the hosted worker guide. Review the model license and provenance before approving an installation. Both paid Positronic plans include model serving and shared organization context; dedicated server charges are separate.
NEBULA 0.6 preview is a standalone CLI for Linux and WSL2. Manage machine identities and fabric membership, connect across networks, and read your hosted organizations and clusters. Kubernetes provisioning is not included in this preview.
Download checksums · Read the CLI guide · Open NEBULA
To connect across networks, pair your machines, enroll them in the same paid Positronic organization, and enable relay access for each machine. Start the owner first, then its members. From the extracted NEBULA folder:
./bin/nebula network connect --no-openKeep this running separately from hosted telemetry or model workers. Your machines make outbound HTTPS connections; they need no public listener or root account.
Connect two machines
Use the same build on both machines and initialize each with a distinct name. Transfer pairing documents through a trusted channel.
Read the pairing instructionsThe owner uses an explicit private IPv4 address. Members connect outbound over mutually authenticated TLS; each resource map is available on that machine’s loopback interface.