---
title: "Deploy DSH Superagent | Proactive Coding Agent for a GitHub Repo"
description: "Mention an agent on a GitHub issue, get a pull request back. Self-hosted."
category: "AI/ML"
url: https://railway.com/deploy/dsh-superagent
---

# Deploy DSH Superagent | Proactive Coding Agent for a GitHub Repo

Mention an agent on a GitHub issue, get a pull request back. Self-hosted.

**[Deploy DSH Superagent | Proactive Coding Agent for a GitHub Repo on Railway](https://railway.com/template/dsh-superagent)**

Machine-readable deploy manifest (JSON, validated by TemplateCI): https://railway.com/deploy/dsh-superagent/manifest.json

- **Creator:** Two Workspaces
- **Category:** AI/ML

## Template content

### dsh-superagent https://raw.githubusercontent.com/two-workspaces/dsh-superagent/main/icon.svg

- **Source:** https://github.com/two-workspaces/dsh-superagent
- **Health check:** /healthz
- **Public domain:** Yes

## Documentation

# Deploy and Host DSH Superagent on Railway

[![Deploy on Railway](https://railway.com/button.svg)](https://railway.com/new/template/dsh-superagent?referralCode=sLjDXb&utm_medium=integration&utm_source=button&utm_campaign=dsh-superagent)

[DSH Superagent](https://github.com/deepseek-ai/dsh-superagent-prototype) is a proactive coding agent for one GitHub repository, built on [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness). Mention its account on an issue, and it opens a ticket, works in its own session and checkout, and answers in the same conversation, usually with a pull request. A web panel shows what needs a person. This template runs it always-on behind a password, with its state on a volume.

This is a community template, not made or supported by DeepSeek. DeepSeek publishes Superagent as a **research prototype for reference, not an official product or production-ready software**, on a preview release of DeepSeek Harness. Try it on a repository you can afford to experiment on.

## About Hosting DSH Superagent

One service, one volume at `/data`:

- DeepSeek Harness and the Superagent bundle, both pinned, listening only inside the container
- Caddy on the public port, with a password in front of the panel, since DeepSeek Harness has no login of its own
- `git`, `gh`, Node 22, Python 3 and build tools, so the agent can run your repo's builds and tests

The volume holds the agent's database, sessions, a clone of your repository and one checkout per ticket.

## Common Use Cases

- Hand off issues by writing `@your-agent fix this` and get a pull request back
- Let an agent pick up small fixes on a repo without keeping your laptop on
- Try DeepSeek's take on an agent that acts only when asked, and speaks only when someone must act

## Dependencies for DSH Superagent Hosting

- A **dedicated GitHub user account** for the agent, with write access to your repository. A GitHub App won't work, because Superagent reads the account's notifications. Don't use your own account: everything the agent writes appears under its name.
- A classic GitHub token for that account with the `repo`, `read:org`, `notifications` and `workflow` scopes
- A model provider key: DeepSeek, or any other provider you add in the panel

### Deployment Dependencies

- [DSH Superagent on GitHub](https://github.com/deepseek-ai/dsh-superagent-prototype)
- [Superagent deployment guide](https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/docs/006-deployment.md)
- [DeepSeek Harness on GitHub](https://github.com/deepseek-ai/deepseek-harness)
- [DeepSeek API keys](https://platform.deepseek.com/api_keys)
- [Create a classic GitHub token](https://github.com/settings/tokens/new?scopes=repo,read:org,notifications,workflow&description=DSH%20Superagent) (signed in as the agent account)

### Implementation Details

**First use:**

1. Create the agent's GitHub account, invite it to your repository with write access, and accept the invitation as that account. In that account's [notification settings](https://github.com/settings/notifications), make sure "Participating, @mentions and custom" includes **On GitHub**. With email only, mentions never reach the agent, and nothing reports an error.
2. Deploy with `AGENT_GITHUB_TOKEN`, `GITHUB_REPOSITORY` (`owner/name`) and `GITHUB_ALLOWED_LOGINS` (your GitHub login; only these people can ask the agent for work).
3. Open your Railway domain and log in as `WEB_USER` with the generated `WEB_PASSWORD` from the Variables tab.
4. If you didn't set `DEEPSEEK_API_KEY`, add a provider under Settings → Models.
5. Choose **Superagent** in the sidebar. Overview names anything still missing.
6. From an allowed account, comment `@your-agent do something small` on an issue. A ticket appears within about a minute.

**Changing the repository or the allowed logins:** edit the variable and redeploy, or change them in the panel under Settings. A redeploy only applies a variable that changed, so edits made in the panel otherwise stand.

**Proactive work** is off by default: in `observe` mode, Superagent records what it would have picked up without doing it. A person merges every pull request unless you set up auto-merge in the Delivery requirements prompt.

Notes and limits:

- The agent runs commands without approval, as root inside this container, with its GitHub token on the volume. Keep real secrets out of the service's variables and give the account access to only what it needs.
- Requests arrive by polling the account's notifications, about once a minute, so no webhook is needed. Adding a mention by editing an existing issue isn't picked up; post a new comment.
- It uses about 350 MB of RAM once set up, and reached 400 MB on a one-file ticket in a tiny repo. That fits a 0.5 GB plan for trying it out, but each ticket adds a session plus whatever your builds and tests use, so real work needs more.
- Feishu is supported upstream but not set up by this template.

## Why Deploy DSH Superagent 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 DSH Superagent 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.
- [Weaviate](https://railway.com/deploy/weaviate-1) — Vector database with hybrid search and an API key-protected API
- [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.

Open this page in a browser: https://railway.com/deploy/dsh-superagent