CLI Reference

Print the CLI version, git commit, and build date.

Print a short banner with credits, the upstream repo, and where to file issues.

Interactive first-time setup wizard.

Create a workspace directory bound to your active tenant node.

List every available top-level resource the CLI supports (deploys, sessions, agents, marketplace items, networks scripts) with a one-line summary each.

Generate a shell-completion script for bash, zsh, fish, or PowerShell.

Health-check the local environment, the active tenant node, and your auth state.

Open the full-screen infrastructure dashboard in your terminal — menu-driven access to deploys, sessions, doctor diagnostics, and AI chat.

Authenticate with browser, device code, API key, or password.

Remove the stored API key, JWT, and active node from ~/.

Show whether the CLI is authenticated, who as, when the JWT expires, and which tenant node it points at.

Print only the authenticated username — handy in CI scripts that key off the user.

Round-trip the stored API key against the server.

Print the active tenant node name and URL.

List every node registered to your account, with reachability and tags.

Re-fetch the node directory from the control plane and update the local cache.

Switch the default tenant node.

Register a self-hosted vxnode container (BYO hardware: Docker / bare-metal / Tailscale Funnel / Cloudflare Tunnel).

Partial-update an existing node record.

Remove a node record from the control plane via DELETE /api/v1/auth/nodes/{id}.

Print the entire current configuration as YAML/JSON.

Show one specific setting by dotted key.

Update a setting.

Print the absolute path to the active config file.

Reset configuration to defaults.

Interactive credential setup wizard.

Deploy a single Docker container by image.

Bundle the current directory (or a git repo) and deploy a FastAPI app on the remote VM.

Build and deploy a React/Vite/CRA app.

Deploy a Next.

Deploy a Django app — provisions Postgres, pgAdmin, runs migrations, then serves via gunicorn.

Deploy a Node.

Deploy a generic Python app — runs the entry script via systemd; pip installs from requirements.

Build a Go binary on the remote VM and run it as a systemd unit (or behind nginx + Traefik in SSL mode).

cargo build --release on the remote VM, then run the resulting binary as a systemd unit (or behind nginx + Traefik in SSL mode).

Build a C++ project (CMake) on the remote VM and run the resulting binary as a systemd unit.

Deploy a PHP application served by nginx + php-fpm.

Deploy a static-site bundle (HTML/CSS/JS, no build step) — uploaded over SSH and served by **host systemd nginx** (NOT the Docker/Traefik path).

Build an Angular app (ng build) on the remote VM and serve the dist/ output via nginx (Traefik in SSL mode).

Build a Vue.

Deploy an Expo (React Native web) app — `npx expo export` then serve the web bundle via nginx + Traefik.

Deploy a Flask app via gunicorn behind nginx (Traefik in SSL mode).

Deploy a Maven/Gradle Java app (executable jar) as a systemd unit, fronted by nginx + Traefik.

Deploy a Laravel app served by nginx + php-fpm; runs composer install, php artisan migrate, and storage:link.

Provision a self-managed PostgreSQL (Metal DB) on a remote VM — adds the user, creates the database, opens the firewall, and writes the resulting connection string into Vault under `metaldb/<resource>`.

Install the Claude Code agent CLI on a remote VM — non-interactive setup of `claude` with API key resolution from Vault (or --api-key).

Install the OpenAI Codex agent CLI on a remote VM — provisions Node, downloads codex, writes the API key.

Install the Hermes agent CLI on a remote VM (a vxcloud-managed minimal devops agent).

Install one of the catalog services on the remote VM.

Print the entire embedded catalog: every service name plus which install modes (baremetal/docker/kubernetes) are supported.

Ship and run an arbitrary local shell script as a remote installer.

Apply a docker-compose.

List every Docker container on the remote host (id, name, status, ports).

Show status (running/exited/created), uptime, image, and exposed ports for a container by name.

Start a stopped Docker container.

Stop a running Docker container.

Stop then start a Docker container.

Stop and remove a Docker container.

Tail journalctl logs for a systemd unit (ollama.

sudo reboot the remote host.

sudo shutdown -h +1 on the remote host.

Run apt autoremove + apt clean + journalctl --vacuum-time=3d on the remote host.

docker system prune -af --volumes on the remote host.

sudo systemctl restart docker on the remote host.

Print free -h plus head of /proc/meminfo.

Print df -hT plus the largest directories under /home, /opt, /var/lib.

List running systemd services on the remote host.

Alias of `services list` — Docker containers on the remote host.

Kill whatever process is bound to <port> on the remote host (sudo fuser -k).

sudo systemctl stop <unit> on the remote host.

List recent provisioning sessions for the active node, with status and duration.

Show details and contents of a specific session — files staged, terraform plan, entrypoint output.

Interactively browse a session’s directory tree.

Edit a file within a session locally (not fully supported remotely — best for tweaking generated terraform files before re-applying).

Re-run a previously planned deploy from a saved session — useful for promoting a dry-run to production without re-typing flags.

Fetch terraform state and generated artifacts for a session — useful for reviewing what was deployed or feeding state into a CI gate.

Tear down a previously provisioned resource.

Check vxcloud platform health or the status of deployed cloud resources (filterable by app/cloud/service).

Import a git repository directly into a session’s application directory — for editing/reviewing without a separate clone.

Multi-language code generation.

Higher-level DevOps orchestration: combines git + docker + VM + pipeline tools to plan and execute a multi-step task.

AI-powered git operations — generate commit messages, summarize PRs, draft branch names from issue text.

Run multiple AI sub-agents in parallel.

List available parallel-agent presets.

Call a specific agent tool directly with no LLM in the loop.

List every tool available to a given agent.

Dashboard summary across every sub-resource (agents, deployments, training, etc.

List configured LLM providers and their default models on the tenant node.

Kafka / Redis / Celery connectivity + active-task counts.

Roll-up health status across every registered model.

Server-side agents (the My Agents tab).

Datasets — list / get / upload (multipart) / download (binary) / preview rows / delete.

Fine-tuning jobs — list / get / create / delete / delete-all.

Training jobs — full CRUD plus clone, restart, logs, metrics, chat, export, tests, qa.

Pipelines (training jobs where type=pipeline).

Model serving — generate a deployable vLLM OpenAI-compatible artifact (docker-compose + command args + test curl) for a base model and optional LoRA adapters.

Vector artifacts produced by the source pipeline — semantic query, FAISS/ChromaDB download, 2D PCA visualize, promote-to-production, upload, delete.

Knowledge bases — list / get / create / delete / delete-all.

Tools & actions — list / create / update (PATCH) / delete.

MCP servers — list / add / refresh / delete.

Evaluation runs (Benchmarks tab).

Programming tab — run a one-shot snippet on the tenant node (python / bash) and persist / fetch saved snippets.

AgentControl-side models (Upload Custom + Soft-Delete All from the Models tab).

Model endpoints (My Deployments sub-tab).

Browse-Models catalog (the Models > Browse Models sub-tab).

Web assets — list / create / delete.

Legacy /benchmarks/ surface.

AgentControl workflow shim (list + trigger).

List repos, import a dataset from a repo, browse repo contents.

Deploy targets — VMs available as deployment hosts.

Discover the agentcontrol endpoints reachable on the tenant node.

Persistent flag on every `agentcontrol` sub-command — overrides X-Tenant-ID for that call.

List every catalog agent with category + availability.

Show the full catalog entry for one agent — description, default port, exposed endpoints, tags, capabilities, env vars.

Deploy a marketplace agent onto a target VM.

List catalog models (FAISS / ChromaDB / sentence-transformers presets).

Show full catalog entry for one model.

Deploy a marketplace model onto a target VM.

List Terraform marketplace templates.

Show a solution template's variables, outputs, and default values.

Provision a marketplace solution.

List every pipeline visible to your tenant, with last-run state.

Create a new pipeline bound to a Git repo.

Show pipeline config + recent builds.

Patch pipeline metadata (branch, workflow file, env vars).

Delete a pipeline.

Trigger a pipeline run — fires a GitHub `workflow_dispatch` for the configured branch/workflow.

List recent builds (optionally scoped to one pipeline).

Show one build's metadata, duration, and step summaries.

Stream a build's logs.

Connect a Git provider — interactive OAuth (recommended) or `--token` for headless/CI use.

List Git provider connections currently usable by pipelines in this workspace.

Disconnect a Git provider — revokes the stored PAT/OAuth token in Vault.

Manage per-pipeline environment variables (injected into build runs).

Run an embedded diagnostic script.

List every diagnostic script available, with description and aliases.

Create a new sandbox.

List your sandboxes — name, image, age, TTL remaining, public URL.

Show one sandbox in detail — bound ports, env, proxy URL, SSH coordinates.

SSH into a sandbox.

Run a one-shot command in a sandbox (no interactive shell).

Push the auto-destroy time out by N minutes.

Tear down a sandbox immediately.

List missions in the current workspace, with last/next run.

Create a one-shot or recurring mission.

Show a mission and its run history.

Force a mission to run now, regardless of schedule.

Pause (or resume) a recurring mission without deleting it.

Delete a mission and its run history.

List recent mission runs (across missions), optionally filtered.

Show runtime health — policy engine, audit sink, gateway, agent registry.

Manage allow/deny policies that gate agent tool invocations.

Tail or query the local audit log.

Inspect / restart the local gateway that proxies agent → tool calls.

Crawl a starting URL with configurable depth and concurrency.

Run a query across one or more search engines.

LLM-orchestrated multi-step research.

Manage individual robots.

Group robots into fleets for bulk commands and shared policies.

Send a typed command to a robot or fleet (e.

Stream live telemetry for a robot or fleet.

Dashboard counters for your organization: contacts, companies, open deals, active sequences, and the email funnel (sent / opened / replied) from the response's email_stats block.

Send one tracked email.

List tracked outbound emails with their engagement state.

Health of the Go email worker on your tenant node.

Search the global prospect pool — people by default, companies with --companies.

Counts beside each filter for a given filter set — seniority, department, country, email status, industry.

Your organization's reveal meter for the current period: used, allowance, remaining.

Un-mask one pool person.

Copy pool rows into your organization's saved leads.

Everything the pool knows about one person, plus your relationship to them — saved yet? revealed? already a contact? Masking still applies: fields you have not paid for are reported as present but withheld, never invented.

One company in the pool with its people split into two tables: new prospects, and the ones already in your CRM.

Your organization's saved leads.

One saved lead, the live pool row behind it, and any drift between the two — your snapshot was taken at save time and the pool moves on.

Update a saved lead's status or notes.

Saved lead → Contact.

Convert many saved leads at once.

Pool → Contact in one action: save, reveal if needed, convert.

Crawl a company's own website and fold what it finds back into the pool — addresses, people, company fields.

Right to be forgotten.

List the organization's saved pool searches.

Save the current filter flags as a named search.

Persistent flag on every `leads` command.

The leads commands name failures by HTTP status rather than by matching message text.

The command group is newer than some checked-in binaries.

Multi-provider AI chat in your terminal.

Docker-container deploys (alias of `vxcli deploy container` with extra Docker-specific helpers).

Provision virtual machines on Alibaba / AWS / Azure / Google Cloud / Linode / DigitalOcean / UpCloud / bare metal.

Provision a managed Kubernetes cluster (EKS / AKS / GKE) or apply a manifest.

Deploy serverless functions (AWS Lambda / Cloud Run / Azure Functions).

Manage cloud object storage (S3 / GCS / Azure Blob).

Configure cloud networking (VPC, subnets, security groups, peering).

Manage InstaNodeBox tenant VMs — short-lived, vxcloud-managed boxes used to dry-run deploys without paying for cloud capacity.

`vxcli deploy fastapi/react/nextjs/django/golang/.

Whatever you pass to `--command` is forwarded to the node, which splits on every `,` and strips quotes before handing it to `docker run`.

`--enable-ssl` puts an HTTP nginx proxy in front of `--port`.

The Docker template healthcheck is `wget --spider http://localhost/health`.

Every stack deploy lands at `/opt/<type>-deploy/bin/docker-compose-<uuid>.

`sebp/elk` runs Elasticsearch + Logstash + Kibana in one container = 3 JVMs.

node1.