--- title: "Deploy Aegra" description: "A self-hosted backend for building and running LangGraph AI agents." category: "AI/ML" url: https://railway.com/deploy/aegra --- # Deploy Aegra A self-hosted backend for building and running LangGraph AI agents. **[Deploy Aegra on Railway](https://railway.com/template/aegra)** - **Creator:** INF Labs - **Category:** AI/ML ## Template content ### Postgres https://devicons.railway.app/i/postgresql.svg - **Image:** ghcr.io/railwayapp-templates/postgres-ssl:18 ### Aegra https://www.aegra.dev/logo.svg - **Source:** codestorm-official/aegra-railway - **Public domain:** Yes ## Documentation # Deploy and Host Aegra on Railway Aegra is an open-source backend for running LangGraph-compatible AI agents and Agent Protocol workloads. This template packages Aegra for Railway with PostgreSQL persistence and a minimal starter graph, providing a clean foundation for self-hosted agent applications. ## About Hosting Aegra This Railway template deploys Aegra together with PostgreSQL as its persistent data layer. Aegra runs as a stateless application service, while PostgreSQL stores persistent application state, metadata, and LangGraph checkpoints. The deployment is designed for developers who want to run LangGraph-compatible agents on their own infrastructure without manually assembling the runtime environment. The Aegra service is built from the dedicated Railway wrapper repository: `codestorm-official/aegra-railway` The repository keeps the Railway deployment lightweight and avoids development-only dependencies and example configurations from the upstream monorepo. ## Common Use Cases - Self-hosted LangGraph agent backends - Agent Protocol-compatible applications - Stateful AI assistants - AI application APIs - Custom LLM agents - Tool-using agents - Multi-step agent workflows - Internal AI services - Backend infrastructure for AI products - Research and experimentation with agent systems ## Dependencies for Aegra Hosting This template uses: - **Aegra** for the agent runtime and API layer - **LangGraph** for graph-based agent execution - **PostgreSQL** for persistent state and checkpoints - **Railway Private Networking** for secure internal database communication - **Railway managed infrastructure** for application deployment and persistence ## Architecture ```text Users / Applications │ ▼ Aegra │ ▼ PostgreSQL ```` Aegra handles the agent runtime and application layer, while PostgreSQL provides persistent storage. ## Aegra Railway Repository The deployment uses a small wrapper repository specifically designed for Railway. Its purpose is to provide: * A clean production container * A minimal Aegra configuration * A starter LangGraph graph * PostgreSQL integration * Reproducible dependency installation * A Railway-friendly runtime * No dependency on upstream development examples This keeps the deployment easier to maintain and reduces unnecessary runtime complexity. ## Starter Graph The template includes a minimal LangGraph smoke-test graph. The starter graph: * Requires no LLM API key * Accepts a simple message * Executes a basic LangGraph flow * Returns the input message * Can be replaced with a real agent implementation Conceptually: ```text START │ ▼ Echo │ ▼ END ``` This gives users a working baseline immediately after deployment while keeping the template provider-neutral. ## Persistent Architecture Aegra itself remains stateless. Persistent application data is stored in PostgreSQL, including runtime metadata and LangGraph checkpoints. | Component | Purpose | | ---------- | -------------------------------------------- | | Aegra | Agent runtime and application backend | | PostgreSQL | Persistent application state and checkpoints | The Aegra service does not require its own persistent volume. ## LangGraph Compatibility Aegra is designed to run LangGraph-compatible applications. Developers can replace the included starter graph with their own graph implementation and use Aegra as the serving layer for: * ReAct agents * Stateful assistants * Tool-calling agents * Long-running workflows * Multi-step reasoning systems * Custom agent architectures ## LLM Provider Flexibility The starter deployment does not require any specific model provider. When replacing the starter graph with a real agent, developers can integrate providers supported by their LangChain or LangGraph application, including: * OpenAI * Anthropic * OpenRouter * Together AI * Self-hosted model endpoints * Other compatible providers This keeps the Aegra runtime independent from a single LLM vendor. ## Scaling The default deployment uses Aegra's local execution mode, which is suitable for a single application instance. The architecture can later be extended for more advanced workloads, including: * Dedicated background execution * Redis-backed coordination * Multiple application replicas * Horizontal scaling * More advanced agent workloads For the base template, the simpler single-instance architecture keeps deployment predictable and lightweight. ## Aegra vs Managed Agent Platforms | Feature | Aegra | Managed Agent Platform | | -------------------------- | ----- | ---------------------- | | Self-hosted | ✅ | ❌ / Limited | | LangGraph compatibility | ✅ | ✅ | | Agent Protocol support | ✅ | Varies | | Infrastructure control | ✅ | ❌ | | PostgreSQL persistence | ✅ | ✅ | | Custom agent code | ✅ | Varies | | Model provider flexibility | ✅ | Varies | | Deployment portability | ✅ | Limited | | Railway hosting | ✅ | ❌ | ## Why Self-Host Aegra? Self-hosting Aegra gives developers greater control over their agent infrastructure. This can be useful for: * Private AI infrastructure * Internal business applications * Custom networking requirements * Independent model selection * Persistent agent state * Integration with internal APIs and services * Reducing dependency on managed agent deployment platforms ## Why Deploy Aegra on Railway? Railway provides the infrastructure needed to run Aegra and PostgreSQL without manually managing servers. Railway handles: * Application deployment * Private networking * Persistent database storage * Automatic HTTPS * Environment management * Container builds * Service orchestration * Infrastructure scaling This allows developers to focus on building agents instead of maintaining the surrounding infrastructure. ## Who Is This Template For? This template is suitable for: * LangGraph developers * AI application developers * Backend engineers * AI platform teams * Researchers building agent systems * Developers building self-hosted AI products * Teams looking for greater control over agent deployment ## Aegra on Railway Aegra and PostgreSQL together provide a compact foundation for self-hosted LangGraph-compatible applications. The included starter graph makes the deployment immediately testable, while the dedicated Railway wrapper repository keeps the infrastructure clean and production-oriented. From there, developers can replace the starter graph with their own agent logic and extend the runtime as their application grows. ## Similar templates - [Chat Chat](https://railway.com/deploy/-WWW5r) — Chat Chat, your own unified chat and search to AI platform. - [stella](https://railway.com/deploy/stella) — Self-host stella with web, API, Postgres, Redis, and object storage. - [Hermes Agent | OpenClaw Alternative with Dashboard](https://railway.com/deploy/hermes-agent-or-openclaw-alternative-wit) — Self-Hosted Hermes AI Agent for Telegram, Discord & Slack Open this page in a browser: https://railway.com/deploy/aegra