Deploy AgenticOS

Deploy and Host AgenticOS on Railway

AgenticOS is an open source platform for building and running AI agents as a team: agents with tools and knowledge bases, chats, routines that run on a schedule, approvals, and connections to Slack, Mattermost, Telegram and more. This template runs AgenticOS 1.27 with an owner account ready to sign in, sign-up locked to invitations, and your uploaded files kept safe across redeploys.

About Hosting AgenticOS

The template deploys four services:

• AgenticOS: the web app you open in your browser
• API: the AgenticOS server. It also runs the background worker (document ingestion, schedules, syncs) and applies database migrations on every start. Uploaded files live on its volume.
• Postgres: Railway's Postgres 16 with pgvector, which AgenticOS uses for knowledge bases
• Redis: a cache for rate limits and live updates

What it does for you:

• You can sign in right away. The owner account is created on first boot with your email and a generated password, and it is the platform admin.
• Strangers cannot register. AgenticOS allows open sign-up by default; this template switches it to invite only on first boot, so your public URL does not admit anyone. You invite your team from inside AgenticOS.
• A working agent from minute one. Paste a model API key at deploy time and the demo agent answers immediately. Skip it and add keys later under Settings.
• Your files survive redeploys. Chat attachments and documents are stored on a volume.
• Chat buttons work. Approve and Reject buttons in Slack or Mattermost call back to the API's public address, which is set for you.
• Pinned to AgenticOS 1.27.0, so a redeploy never pulls a surprise upgrade.

Common Use Cases

• An internal AI assistant for a team, with its own knowledge base of your documents
• Agents that answer in Slack, Mattermost or Telegram, with approvals before risky actions
• Scheduled routines: weekly reports, inbox triage, research digests
• A shared place to build, test and share agents across a company, with budgets per team

Dependencies for AgenticOS Hosting

• PostgreSQL 16 with pgvector, on a Railway volume
• Redis 8
• The official ghcr.io/vstorm-co/agenticos-backend and agenticos-frontend images

Deployment Dependencies

• AgenticOS on GitHub
• AgenticOS configuration reference
• An API key from a model provider (Anthropic, OpenAI, Google or OpenRouter)

Implementation Details

At deploy time you are asked for one thing: ADMIN_EMAIL, the email you sign in with. Optionally, BOOTSTRAP_API_KEY with a key for BOOTSTRAP_PROVIDER (anthropic by default) so the demo agent works straight away. Everything else is generated.

After deploying:

1. Open the API service, go to Variables, and copy ADMIN_PASSWORD.
2. Open the AgenticOS service's URL and sign in with your email and that password. Change the password under your profile.
3. If you skipped the API key, add one under Settings → AI providers.
4. Invite your team under Organization → Members. Sign-up is invite only; to change it, go to Admin → System.

First boot takes a minute or two: the API applies the database migrations, then creates your account, organization and demo agent. The web app shows its sign-in page as soon as the API is up.

Keep VAULT_MASTER_KEY. It encrypts the API keys and tokens AgenticOS stores. If it is lost or changed, those secrets cannot be read and have to be entered again.

Optional extras, add them to the API service any time:

• Email (invitations and notifications): SMTP_HOST, SMTP_PORT, SMTP_USER, SMTP_PASSWORD, EMAIL_FROM
• Google or GitHub sign-in: the OAuth variables in the configuration reference, then set OAUTH_PROVIDERS on the AgenticOS service (for example google)
• Custom domains: add one to each of the two public services, then update PUBLIC_BASE_URL, PUBLIC_API_URL and PUBLIC_WS_URL (API address) and FRONTEND_URL, PUBLIC_SITE_URL and CORS_ORIGINS (web address) on the API, and the matching PUBLIC_* variables on the AgenticOS service

What is not included: AgenticOS's code sandbox (sandboxd), which lets agents run code in throwaway containers. It needs direct access to a Docker daemon, which Railway does not give a service. Agents work fully without it; to add code execution, register a remote sandbox under Connections → Sandboxes.

Scaling: the API runs 2 worker processes (UVICORN_WORKERS), about 450 MB each, which suits a team of ten or so. Raise it for a busy deployment, together with the service's memory.

Troubleshooting:

• The web app shows an error right after deploying: the API is still applying migrations on first boot. Wait a minute and reload.
• The demo agent says it has no model: add a provider key under Settings → AI providers.
• Slack or Mattermost buttons do nothing: check that PUBLIC_BASE_URL on the API is its public https:// address.

Why Deploy AgenticOS on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying AgenticOS on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.