Salus
For builders — human and agent

Build with anything. Ship as yours.

Built it with Claude, Cursor, Replit — or your own editor? Salus takes real code from any builder, human or agent, and runs it as your own governed workload: deploys, preview environments, logs, costs and on-demand GPU. Any framework, no waiting on anyone.

your agent · claude codesalus · mcp

Deploy my expense tracker, please.

Done — it went out through your company’s governed lane.

build ✓security scan ✓policy ✓live

every step checked against IT’s rules · audited

Is it safe? Run a security scan.

Just did — 12 policies checked, nothing critical found.

It feels slow this morning. Can you check the logs?

Found it: the database connection pool is maxed out. I can raise it and redeploy.

Rather roll back to yesterday’s version for now.

Rolled back — yesterday’s version is live again. IT can see everything I did in the audit trail.

rollback ✓audited

API

payments-api

live
p99 24ms · 3 regions

GPU

model-train-a100

scaling
GPU 78% · a100 ×4

Worker

ingest-worker

building
queue 142

Web

dashboard-pr-482

live
preview env
For every team

What will your team build?

Every team has apps they wish they had — the spreadsheet that should be a tool, the workflow stuck in email. With AI, they can build them. Salus is where those apps run: governed, visible, and in your company's own cloud.

Finance

  • Spend & budget dashboards
  • Approval workflows
  • Invoice reconciliation tools

Marketing & Sales

  • Campaign microsites & landing pages
  • Lead-routing and scoring apps
  • Content and asset tools

Operations

  • Internal admin tools
  • Inventory & logistics trackers
  • Process automations

Product

  • Internal dashboards
  • Feedback & research portals
  • Working prototypes

Customer Support

  • Ticket-triage helpers
  • Customer lookup apps
  • Knowledge tools

Data & Analytics

  • Self-serve reporting apps
  • ML-output dashboards
  • Data-entry & cleanup tools
CLI & MCP

Your agent already knows how to drive it.

Each command does one thing and prints what it did. And every one of them doubles as an MCP tool your agents can call — under the same RBAC and guardrails as you.

salus init

Detect your stack and scaffold a Salus project.

salus deploy

Build, scan and deploy in one step.

salus logs api

Stream structured logs by workload.

salus cost workload api

Per-workload cost in real time.

salus scale gpu-worker --gpu a100

Burst GPU on demand.

salus rollback api

Roll back to the last good revision.

Built for the way you work

Everything you need to go from idea to production.

Zero-config deploys

Push a Dockerfile, a Node app or a static site. Salus detects, builds and deploys.

Preview environments

Every PR gets an isolated preview with its own database, secrets and URL.

Logs, traces and metrics

Built-in observability — no extra agent, no extra integration to wire up.

GPU on demand

Run GPU jobs and inference workloads with a single flag. Pay for what you use.

Secrets & networking

Per-environment secrets and private networking, governed by org policy.

MCP for agents

Every operational step is an agent-callable tool — deploy, rollback, logs, databases — using the access of the builder who invoked it.

From commit to production

From 'it works on my laptop' to live — one path.

Salus runs every change through the same flow — so you can move fast without losing control.

  1. step 1

    git push

    commit detected

  2. step 2

    build

    image salus.app/api:9f3c

  3. step 3

    scan

    no critical issues

  4. step 4

    policy

    release window OK

  5. step 5

    deploy

    rollout 100%

  6. step 6

    observe

    logs · traces · cost

A software engineer?

There's a track for you too — real production, pipelines, any framework, your own cloud, and agents as first-class teammates over MCP.

See the engineering track

Ship the AI apps you build — without waiting in the engineering queue.

Sign up and deploy your first app in under five minutes. No credit card required.