WK·001
Alph
An open source CLI for MCP setup, built because editing the same JSON in five different places is a poor use of anyone's evening.
52 commits · 8 active weeks
latest 272c3b3 · 22 Jan 2026
Overview
Alph is a small developer tool with a very ordinary origin: I was tired of manually configuring MCP servers for every coding assistant I used. Each tool had its own config path, its own shape, and its own way of failing quietly when I got one line wrong.
So I built a CLI that could add MCP servers across supported tools, show the change before writing it, keep backups, and remove the setup cleanly later. It was not trying to be a platform. It was trying to remove a brittle step from a workflow that was already useful once you got past the setup.
The reconstruction
WK·001 · alph setup, simulated · no files on your machine are modified
What each step is doing
Context
Alph came out of AskHuman. To use a hosted MCP endpoint, local agents needed to know where it lived. That meant opening config files by hand, making the same change in different formats, and then remembering what I had touched when something broke.
The first user was me, which is usually a decent way to find annoying infrastructure work. If it irritated me, it was going to stop someone newer from trying MCP at all. The fix did not need to be clever. It needed to be safe, repeatable, and boring in the right places.
What changed
A tool that edits local configuration has to earn trust before it writes anything. Alph previews changes, masks secrets, creates a backup, validates the result, and leaves a remove path. That is the part I care about most. The actual JSON editing is not impressive. Making it safe enough to run on a machine you depend on is the work.
New AI tooling often makes the exciting part easy to imagine and the boring setup easy to ignore. That is a bad bargain. If the first step is fragile, people either give up or copy something from a README they only half understand. Alph existed to make the first step less fragile.
The build
The useful work sat around the edges: tool detection, schema validation, backup handling, secret masking, and safe writes. Detection only reads. Writes go through a predictable path. Existing config is preserved rather than casually overwritten.
None of this is glamorous, which is why it matters. A setup tool succeeds when the user does not have to think about it again.
Notes from the archive
Alph did not fail. It did what it was built to do, and it is still open source. What changed is the ecosystem around it. Claude Code, Codex and other tools began adding their own MCP setup commands, in-app flows, and marketplaces. The problem did not disappear, but the platforms moved closer to it.
That made continuing Alph as a standalone product less useful. Slightly annoying, but also a useful signal. If the tools absorb the workflow you were trying to smooth over, you were probably looking in the right direction.
Evidence
| repo | github.com/Aqualia/Alph | public |
| external-link | @aqualia/alph-cli on npm | npm |
| commit | Latest reviewed main-branch commit · 1aea6d9 | 2025-09-22 |
| note | ARCHITECTURE.md, SECURITY.md, SolutionDesign | in repo |
| video | demo-alph.gif recorded CLI run | in repo |
Technical detail
Other exhibits
| WK·002 | AskHuman | A hosted MCP loop for agents that need a human decision instead of another confident guess. | Decision loop |
| WK·003 | Rhya | A live wellness site for a real business, built close enough to the stakeholder that vague feedback was never going to survive long. | Capture |
| WK·004 | Gradience | An assisted grading workflow that was probably right about the direction, but early for the market and the models. | Timeline |
| WK·005 | WatstheStory | A personalised audio briefing system for WhatsApp, built because a short spoken update sounded better than opening ten feeds before breakfast. | Live |