My taskbar used to look like a museum of half-open tools: a SQL client for Postgres, another window for Redis, FileZilla for deploys, PuTTY for the box, and a browser tab for S3. Same day, same product, five password prompts. Nexoxa Bridge started as a selfish fix for that mess — a Windows workspace where PostgreSQL, Redis, SSH, and S3 share one encrypted vault.
The problem was not “I need another SQL editor”
Postgres GUIs are solved enough. The pain was context switching. I would fix a row in production, check a cache key, SCP a config, then SSH in to restart a container. Each hop meant another app, another saved password, another mental model. I did not want a lowest-common-denominator “universal grid.” I wanted purpose-built panels that still lived in one shell.
Constraint #1: credentials stay on the machine
I refused to build a hosted “AI ops” product that asks for production URLs. Connection strings are secrets. So Bridge is local-first: the desktop app encrypts URLs, SSH keys, and S3 credentials in an OS-backed vault (AES-256-GCM). Account login for billing is separate from the vault. If the laptop is offline, the studio still works.
Constraint #2: one URL, many engines
Paste postgresql://, redis://, ssh://, s3://, sftp://, smtp:// — no driver ceremony. Under the hood Bridge is a Tauri shell with an embedded Node server and a TanStack UI. That split let me ship Windows Store builds while iterating on schema browsers, transfer queues, and SSH dashboards without rewriting the whole stack.
What “one workspace” actually means
PostgreSQL gets a schema tree, data grid, Monaco SQL, and an FK knowledge graph. Redis gets a key browser beside those same saved connections. SSH opens a server dashboard — processes, ports, Docker — plus a terminal. S3 and SFTP share a dual-pane transfer studio. SMTP sits in Mail studio for verify-and-send tests. Same sidebar. Same vault. Same session.
AI without pasting production URLs into chat
Copilot is BYOK (bring your own OpenRouter key). For Cursor, I shipped nexoxa-mcp: the agent connects by saved connection name only. The local Bridge process decrypts secrets; the model never needs the raw URL to start a session. That pattern mattered more to me than shipping yet another cloud chatbot glued to a database.
What I deliberately did not build first
macOS/Linux ports can wait. Exotic JDBC engines can wait. A perfect pgAdmin replacement for every admin wizard can wait. The v1 bet was Windows developers who already juggle Postgres + Redis + a VPS — and who will trade a fifth app for one honest free tier (five saved connections on the Microsoft Store).
Shipping and what is next
Bridge is on the Microsoft Store and as an NSIS installer. Pro unlocks the knowledge graph and higher limits at $9/mo. Public packages like nexoxa-mcp are there for automation. If this story matches your taskbar, download it, move five connections into the vault, and see whether you still open FileZilla the same week.
