Can Synra use Railway's private DATABASE_URL (railway.internal)?

No. Private networking only works between services inside the same Railway project. Synra runs outside Railway, so you need the public TCP proxy host and port (proxy.rlwy.net-style), not *.railway.internal.

Where do I find the public Postgres URL on Railway?

Open the Postgres service → Variables (or the Connect / public networking UI). Use the connection string built from the TCP proxy domain and port — Railway enables a TCP proxy for Postgres by default. Internal-only hostnames will fail Synra's connection test.

Should I connect Cursor to my Railway production environment?

Prefer a staging environment (or a separate staging Postgres) first. Railway environments isolate variables and data; mirror that in Synra with separately named connections and endpoints.

Does this work with Cursor Composer and Agent?

Yes. Once the MCP server is in mcp.json and green in Settings → MCP, chat, Composer, and Agent share the same tools.

Can Cursor write to Railway Postgres through Synra?

Not with the default setup. INSERT, UPDATE, DELETE, DROP, ALTER, and TRUNCATE are blocked at the gateway.

Is this different from Connect Cursor to PostgreSQL?

Same gateway and Cursor steps. This post is Railway-specific: public vs private URL, TCP proxy, and environment naming. Generic hosts: Connect Cursor to PostgreSQL.

Connect Cursor to Railway Postgres | Cursor MCP Setup

Railway often hosts the Postgres next to the app. Cursor is where you write migrations and API code. Without a live link, the agent invents columns for a database it never saw.

This walkthrough connects Cursor to Railway Postgres over MCP. The non-generic bits: public TCP proxy vs private railway.internal, Railway environments, and naming connections so Composer doesn't hit production by accident.

Other hosts: Connect Cursor to PostgreSQL. Neon: Connect Cursor to Neon. New to MCP? What is an MCP server.

What you can ask once it's connected

With Railway Postgres behind MCP:

"Compare this migration to the users table on staging — will it fail on existing nulls?"
"How many subscriptions rows have status = 'trialing' older than 14 days?"
"Our seed assumes three orgs. What's actually in this Railway DB right now?"
"List indexes on orders — anything covering (org_id, created_at)?"

A one-off railway connect dump in chat goes stale the next deploy. Live queries don't.

The Railway gotcha: private vs public URL

Inside a Railway project, services talk over private networking — hostnames like Postgres.railway.internal. Your app's DATABASE_URL often points there. That's correct for the app.

Synra is not inside your Railway project. It can't resolve *.railway.internal. If you paste the private URL into Synra, the connection test fails (or hangs) and Cursor never gets tools.

You need the public path Railway exposes via TCP proxy (enabled by default on Postgres). The host looks like something.proxy.rlwy.net with a high port — not 5432 on an internal name.

| URL style | Example shape | Use with Synra? | |-----------|---------------|-----------------| | Private | postgresql://…@Postgres.railway.internal:5432/… | No | | Public / TCP proxy | postgresql://…@….proxy.rlwy.net:15xxx/… | Yes | | App variable that still embeds .railway.internal | whatever your web service references | Only if the host is public — check the hostname |

Railway bills network egress when traffic leaves via the proxy. AI query volume is usually small next to app traffic — still a reason not to let an agent hammer a huge prod table.

Step 1 — Copy the public Postgres connection string

Open your project on railway.com
Select the Postgres service (not your web/API service)
Open Variables (or the Connect / networking UI for that service)
Find the public connection string — or build one from:
- PGUSER / POSTGRES_USER
- PGPASSWORD / POSTGRES_PASSWORD
- PGDATABASE
- TCP proxy host + port (RAILWAY_TCP_PROXY_DOMAIN + RAILWAY_TCP_PROXY_PORT on the Postgres service)

Final shape:

postgresql://postgres:PASSWORD@HOST.proxy.rlwy.net:PORT/railway

Confirm the host is proxy.rlwy.net (or your custom TCP hostname), not railway.internal.

Pick the environment first. Railway environments (e.g. production vs staging) each have their own Postgres variables and data. Prefer staging for Cursor. Same habit as staging-first.

Step 2 — Save it in Synra

Sign in at mcpserver.design (7-day Solo trial is enough)
Connections → Add new connection
Choose PostgreSQL
Paste the public URI
Name it after the Railway environment — Railway staging, Railway production replica — not a vague Postgres
Test connection → Save

If the test fails:

You pasted the private URL → switch to TCP proxy
Password special characters need URL-encoding in the URI
TCP proxy disabled on that service → enable it under the Postgres service networking settings

Credentials stay encrypted in Synra. Cursor only ever sees the MCP endpoint URL.

Step 3 — Copy the MCP endpoint

Endpoints tab → URL like:

https://app.mcpserver.design/api/mcp/aB3xY7zQ9mNpLkRv2wTs

Treat the token like a password. Rotate if it leaks.

Step 4 — Add it to Cursor

~/.cursor/mcp.json (global) or .cursor/mcp.json (per project). Don't commit the live token.

{
  "mcpServers": {
    "synra-railway-staging": {
      "url": "https://app.mcpserver.design/api/mcp/aB3xY7zQ9mNpLkRv2wTs"
    }
  }
}

Name the key after the environment. Second environment later → second Synra connection + second mcp.json entry.

Older Cursor without "url" support:

{
  "mcpServers": {
    "synra-railway-staging": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://app.mcpserver.design/api/mcp/aB3xY7zQ9mNpLkRv2wTs"]
    }
  }
}

Step 5 — Restart and verify

Quit Cursor fully (Cmd+Q on Mac), reopen:

Settings → MCP → synra-railway-staging green
Tools listed: list_tables, describe_table, query_table, execute_sql
Chat (Cmd+L): "List the tables in this database." / "Describe users." / "How many rows in orders?"

Then open Synra Usage and confirm the tool calls landed on the credential you named — what the audit log is for.

Habits that matter on Railway

One Synra connection per Railway environment. If you repoint the URI from staging variables to production, rename the connection (or make a new one). Don't keep calling it staging.

App DATABASE_URL ≠ what Synra needs. Private networking stays for services in the project. Synra always gets the public proxy string.

Password rotate on Railway. Update the connection in Synra and re-test. The MCP endpoint URL is unchanged unless you rotate that separately.

SELECT-only role when you can. Synra blocks writes at the gateway; a least-privilege Postgres user is the second layer. Why read-only matters.

Troubleshooting

Synra test fails, psql from your laptop works. You're probably using a different string locally (public) than you pasted (private). Recopy the proxy URL.

could not translate host name "….railway.internal". Private URL in Synra — replace with TCP proxy host.

Green in Synra, red in Cursor. Bad mcp.json (trailing comma) or stale endpoint. cat ~/.cursor/mcp.json | jq ., then recopy from Endpoints.

Empty / wrong tables. Wrong Railway environment selected when you copied variables. Fix the URI in Synra; don't keep editing Cursor for this.

Agent ignores the DB. Type @ in chat and pick synra-railway-staging once.

More: MCP troubleshooting.

Public Railway URI into Synra, endpoint into mcp.json, then ask something you already know the answer to and check the number.

7-day trial at mcpserver.design — one staging connection is enough.

Related reading:

Connect Cursor to PostgreSQL — any Postgres host
Connect Cursor to Neon — branch-aware Neon setup
Connect Cursor to Supabase — Supabase credentials path
What Your AI Query Audit Log Is For — week-one Usage habits
Staging First — environment ladder
Ask Cursor About Your Database — why live beats paste
MCP Troubleshooting — tools missing
What is an MCP Server? — protocol basics