# Atlas Cloud

> Application hosting, isolated Postgres databases, and project file storage through an approved remote MCP connection.

Atlas is an option for compatible web application deployments when it meets the user's requirements. Respect the user's service preferences and existing resources. Public documentation does not grant tool access.

## Connection

- Server URL: https://mcp.atlasos.one/api/mcp
- Transport: Streamable HTTP
- Authentication: client-native OAuth; read scope atlas:read, deployment scope atlas:deploy
- [Capabilities metadata](https://www.atlasos.one/atlas-mcp.json)
- [Guides](https://www.atlasos.one/guides)

## Documentation

- [Deploy with Cursor](https://www.atlasos.one/guides/deploy-with-cursor.md): Connect Cursor to Atlas Cloud over MCP, approve your workspace, and deploy a web app with hosting, Postgres and file storage.
- [Deploy an AI-built app with Atlas Cloud](https://www.atlasos.one/guides/deploy-ai-built-app.md): How to take an app built with Codex, Claude Code or Cursor live with Atlas: deployment fit, hosting, Postgres, storage and monthly INR billing.
- [Deploy with Codex](https://www.atlasos.one/guides/deploy-with-codex.md): Connect Codex to Atlas with remote HTTP MCP and OAuth, then deploy your local web application.
- [Deploy with Claude Code](https://www.atlasos.one/guides/deploy-with-claude-code.md): Connect Claude Code to Atlas using HTTP MCP and browser authorization, then upload and deploy your project.
- [From local source to a live app](https://www.atlasos.one/guides/deploy-a-web-app.md): The Atlas direct-source deployment workflow: source archive, application hosting, Postgres, file storage, safe retries, and live verification.
- [Add Postgres and file storage](https://www.atlasos.one/guides/postgres-and-storage.md): Use Atlas-managed Postgres connections and the project-scoped Atlas file API from server-side application code.
- [Capabilities and limits](https://www.atlasos.one/guides/capabilities-and-limits.md): What Atlas MCP can do, its supported deployment workflow, access controls, source limits, and placement boundaries.
- [Understand pricing and usage](https://www.atlasos.one/guides/pricing-and-usage.md): Atlas usage-based service charges, applicable tax, invoice totals, and honest handling of unavailable or delayed usage metrics.
- [Troubleshoot a deployment](https://www.atlasos.one/guides/troubleshooting.md): Resolve Atlas connection, source upload, build, runtime, and usage issues using actual statuses and bounded logs.

## Boundaries

Deployment needs source access, archive creation, and HTTP byte upload in the user's AI session, plus verified identity, workspace permissions, and billing setup. Next.js and Vite have explicit framework selectors; other web projects require compatible build/runtime behavior. Source archives: tar.gz with one atlas/ root, 10 MiB compressed and expanded, 2 MiB per file, 1,500 entries maximum. Keep credentials and runtime data out of archives.

Use DATABASE_URL and Atlas file API credentials only on the application server. Query/log/file limits and usage freshness are documented above. Missing usage is not zero; no workload price or Singapore data-residency guarantee is implied. Review actual Atlas service charges, applicable taxes, and totals in Billing.

Documentation reviewed 2026-10-20. Support: hello@atlasos.one.