--- title: "Deploy Mastra Factory" description: "Agent software factory: turn GitHub issues into reviewed PRs with AI agents" category: "AI/ML" url: https://railway.com/deploy/mastra-factory --- # Deploy Mastra Factory Agent software factory: turn GitHub issues into reviewed PRs with AI agents **[Deploy Mastra Factory on Railway](https://railway.com/template/mastra-factory)** Machine-readable deploy manifest (JSON, validated by TemplateCI): https://railway.com/deploy/mastra-factory/manifest.json - **Creator:** Auromations - **Category:** AI/ML - **Total deploys:** 1 ## Template content ### factory https://mastra.ai/favicon/new-brand/icon-512.png - **Image:** ghcr.io/hmseeb/mastra-factory-railway:0.24.0-r1 - **Health check:** /health - **Public domain:** Yes ### postgres https://devicons.railway.app/i/postgresql.svg - **Image:** pgvector/pgvector:pg18 - **Start command:** `/bin/sh -c "unset PGPORT; exec docker-entrypoint.sh postgres --port=5432"` ## Documentation # Deploy and Host Mastra Factory on Railway [Mastra Factory](https://mastra.ai/factory) is an open-source software factory from the team behind the Mastra agent framework. Connect a GitHub repository and Factory turns issues into plans, implementations and pull requests, with coding agents doing the work and your team approving each stage: intake, triage, planning, building, review and done. ## About Hosting Mastra Factory Factory is a Node.js server that serves its own web UI and API, keeps all state in Postgres with pgvector, and runs agent commands in a sandbox. This template runs the upstream Factory project (`mastra-ai/softwarefactory-template`, `@mastra/factory` 0.24.0) on Railway next to a pgvector Postgres, with no Mastra platform account needed. Agents run their commands in a local sandbox inside the Factory container: they clone your repository onto a Railway volume, install, test and push from there. Set an E2B key later if you want every session in an isolated VM instead. Sign-in goes through WorkOS, and only the email addresses you list can get in. Anyone else who signs in is refused, so strangers cannot use your container to run code. ## Common Use Cases - Turn GitHub issues into reviewed pull requests, with a person approving the plan and the merge - Hand a backlog of small fixes, refactors and test gaps to coding agents - Run Claude, OpenAI or other models on your own subscriptions or API keys, with sessions you can watch live - Give a small team one shared agent workspace with per-person model accounts ## Dependencies for Mastra Factory Hosting - Postgres with pgvector (included) for agent memory, sessions and Factory data - A free WorkOS account for sign-in - A GitHub App you create on your account or organization, for repository access - A model account: Anthropic or OpenAI subscription sign-in, or an API key from any supported provider ### Deployment Dependencies - Mastra Factory: https://mastra.ai/factory and docs at https://factory.mastra.ai - Upstream source: https://github.com/mastra-ai/softwarefactory-template - Container image for this template: https://github.com/hmseeb/mastra-factory-railway - WorkOS: https://dashboard.workos.com - GitHub App permissions Factory needs: https://factory.mastra.ai/configure/github ### Implementation Details **Before you click Deploy (5 minutes).** The form asks for three values. 1. Create a WorkOS account at https://dashboard.workos.com. Under **Authentication**, turn on **Email + Magic Auth** (or whichever sign-in method you want). Copy the **API key** (`sk_...`) and **Client ID** (`client_...`) from **API Keys**. 2. Paste them into `WORKOS_API_KEY` and `WORKOS_CLIENT_ID`. 3. Put your email in `FACTORY_ALLOWED_EMAILS`. Add teammates separated by commas. Only these addresses can sign in, and only once WorkOS has verified them. **Right after it deploys.** In WorkOS, go to **Redirects** and add `https:///auth/callback`, using the domain Railway gave the `factory` service. Then open your domain and sign in. (The button on the sign-in page reads "Continue with GitHub", but it opens WorkOS.) **Connect GitHub.** Factory needs a GitHub App to see your repositories. Create one at **GitHub → Settings → Developer settings → GitHub Apps → New GitHub App**: - Callback URL and Setup URL: `https:///auth/github/callback` - Webhook URL: `https:///web/github/webhook`. Webhook secret: copy `GITHUB_APP_WEBHOOK_SECRET` from the `factory` service's Variables tab. - Repository permissions: **Contents**, **Issues** and **Pull requests** set to Read and write. Metadata is read-only and added automatically. - Events: Issues, Issue comment, Pull request, Pull request review, Pull request review comment. After creating it, generate a private key, then set these five variables on the `factory` service and redeploy: `GITHUB_APP_ID`, `GITHUB_APP_SLUG` (the name in the app's URL), `GITHUB_APP_CLIENT_ID`, `GITHUB_APP_CLIENT_SECRET`, and `GITHUB_APP_PRIVATE_KEY` (paste the whole `.pem` file). In Factory's onboarding, **Connect GitHub**, install the app on the repositories you want, and pick one. **Choose a model.** Onboarding offers sign-in with an Anthropic (Claude Pro or Max) or OpenAI (ChatGPT) subscription, or an API key from a long list of providers. Each person can use their own account. For ChatGPT sign-in, first turn on device code sign-in under ChatGPT **Settings → Security**. **Pull requests from agents.** GitHub App tokens cannot run `gh` commands inside the sandbox. To let agents open pull requests, save a classic personal access token with `repo` scope under Factory's **Settings → Repositories → GitHub CLI tokens**. **Variables** | Variable | What it does | | --- | --- | | `WORKOS_API_KEY`, `WORKOS_CLIENT_ID` | WorkOS sign-in | | `FACTORY_ALLOWED_EMAILS` | Comma-separated emails allowed in. Required; the server refuses to start without it | | `MASTRACODE_PUBLIC_URL` | Public origin for every OAuth callback. Change to `https://your.domain` after adding a custom domain, and update WorkOS and GitHub to match | | `FACTORY_CREDENTIAL_ENCRYPTION_KEY` | Encrypts stored model keys and tokens, generated per deploy. Never change it, or stored credentials become unreadable | | `WORKOS_COOKIE_PASSWORD` | Seals sessions and signs OAuth state, generated per deploy | | `GITHUB_APP_*` | Your GitHub App. All five must be set for GitHub to light up | | `GITHUB_APP_WEBHOOK_SECRET` | Generated per deploy. Paste the same value into your GitHub App | | `FACTORY_SANDBOX_PROVIDER` | `local` runs agent commands in the Factory container | | `E2B_API_KEY` | Optional. To use isolated E2B sandboxes, set this and clear `FACTORY_SANDBOX_PROVIDER` | Other upstream settings (Linear, Jira, GitLab, Slack, Amazon Bedrock) can be added as variables on the `factory` service; see the [environment reference](https://factory.mastra.ai). **Troubleshooting** - *WorkOS shows "Invalid redirect URI".* Add `https:///auth/callback` under WorkOS **Redirects**, exactly, including `https://`. - *Sign-in finishes but lands back on the sign-in page.* Your email is not in `FACTORY_ALLOWED_EMAILS`, or WorkOS has not verified it. Fix the variable and redeploy. - *Deploy fails with `FACTORY_ALLOWED_EMAILS is required`.* Set it; this is the lock that keeps the instance private. - *Onboarding says GitHub is "Unavailable for this deployment".* One of the five `GITHUB_APP_*` variables is missing. `https:///web/github/status` lists which. - *"Couldn't load GitLab / Jira / incident.io".* Expected when those integrations are not configured. Click **Skip for now**. **Scaling and limits.** This template runs one Factory replica. Repository checkouts and agent work share the container's CPU and memory, so give the `factory` service more resources if you run many sessions at once, or move to E2B. Running several replicas needs Redis (`REDIS_URL`) and is not set up here. Data lives in Postgres, and checkouts live on the `/data` volume, so both survive redeploys. ## Why Deploy Mastra Factory 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 Mastra Factory 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. ## Similar templates - [Chat Chat](https://railway.com/deploy/-WWW5r) — Chat Chat, your own unified chat and search to AI platform. - [Möbius v0.1 Personal AI OS](https://railway.com/deploy/mobius-v01-personal-ai-os) — Self-hosted AI agent that builds and runs apps for you, with memory. - [stella](https://railway.com/deploy/stella) — Self-host stella with web, API, Postgres, Redis, and object storage. Open this page in a browser: https://railway.com/deploy/mastra-factory