Deploy and Host IronClaw on Railway

For a private Rust-based AI agent workspace on Railway, use IronClaw. This template packages NEAR AI's native web interface with generated access credentials, a persistent runtime home and a workspace volume. Configure the model provider before running AI tasks. It fits a single-tenant workspace for conversations, tools, memory and automations; it is not a shared sandbox for unrelated customers.

About Hosting IronClaw

The official nearaidev/ironclaw image is pinned by digest. The hosted single-tenant volume profile stores its runtime home under /data/ironclaw-reborn, including configuration and durable state. The workspace root is also inside the same persistent Railway volume. One replica is used for this single-writer volume layout.

Open the generated HTTPS domain. When prompted for access, use IRONCLAW_REBORN_WEBUI_TOKEN from the service variables. IRONCLAW_REBORN_SECRET_MASTER_KEY is generated independently; keep it unchanged and back it up separately from your data volume. Provider setup is available in the native UI: no LLM key is required merely to start the service, but AI runs need a configured provider and model.

Common Use Cases

Private AI conversations and persistent agent workspaces.
Configurable LLM providers and models in the web interface.
Memory and scheduled agent automations.
MCP and other upstream extensions, subject to their individual configuration and runtime requirements.

Dependencies for IronClaw Hosting

One Railway service using the pinned official IronClaw container.
A persistent /data volume and one replica.
A supported LLM provider and its credentials to run AI tasks.
Optional integration credentials for extensions you explicitly install and configure.

Deployment Dependencies

Railway HTTPS routing and one persistent volume. The model provider is configured separately after deployment; keep access secrets stable across restarts.

Implementation Details

The native HTTP server listens on 0.0.0.0:3000; Railway checks /api/health. The profile is hosted-single-tenant-volume, with IRONCLAW_REBORN_HOME=/data/ironclaw-reborn and IRONCLAW_REBORN_WORKSPACE_ROOT=/data/ironclaw-reborn/workspace. The WebUI identity matches the profile's runtime owner (reborn-cli). The official entrypoint manages volume ownership and runs the application under its unprivileged runtime account.

This template does not provision Docker-in-Docker, a privileged container, or a Railway Sandbox API token. Features requiring an additional runtime must be configured separately; a normal Railway container does not provide every optional upstream execution lane.

Frequently Asked Questions

Can I add my model key after deployment? Yes. Open the private UI and configure a provider. Startup readiness does not guarantee that a provider account has credits or that every model supports every tool.

What survives a restart? The attached volume keeps the runtime home, configuration, durable conversations and workspace files. Keep both the volume and secret master key stable.

Does the template require PostgreSQL? This layout uses the upstream volume profile. PostgreSQL-backed production profiles are separate deployment configurations, not required for this template.

Is this an official template from NEAR AI? It is a community Railway template using the official IronClaw image.

Why Deploy IronClaw on Railway?

Railway supplies a generated HTTPS domain, managed variables, persistent volumes, health checks and logs. This template adds ready-to-use access secrets, documented storage paths and a profile designed for a persistent single-tenant deployment.

Source and licenses: nearai/ironclaw, MIT and Apache-2.0 as specified upstream.