Question 1

Which MCP clients are actually supported?

Accepted Answer

Not documented, and the listing does not settle it. There is no README, no clients directory, and nothing beyond package.json in the visible root. What you can see is skills wiring for three agent setups — .agents/skills, .claude/skills and .pi/skills — plus a top-level typesafe-ai directory. That is a strong hint about intended hosts, not a statement of compatibility.

Question 2

What do noul, choice and score return, and do I need all three?

Accepted Answer

The About line gives you the three names and nothing else. No tool schema, parameter list, response type or scoring range appears anywhere on the page, and the contents of src are not shown. Whether a client can call one judgment independently or whether the three are coupled into a single request is undocumented, so budget time to read src before wiring this into an agent.

Question 3

What config does it need — an API key, a local model, an endpoint?

Accepted Answer

There is a .env.example, which confirms external configuration is expected, but the page shows neither its contents nor the variable names. So you cannot tell whether you point this at a hosted TypeSafe endpoint, run something locally, or supply credentials, and you cannot see whether .gitignore already protects whatever you add. Check .env.example and src yourself before assuming a quick local run.

Question 4

Is this a demo or something to build a service on?

Accepted Answer

The name says demo and the signals agree: a single commit, zero stars, no forks, no issues, no releases and no tags. That means no stability guarantees, no changelog and no outside user who has already hit the failure you are about to hit. Worth reading as a reference for how someone exposed System One judgments over MCP, not as a dependency to pin.

Question 5

What is actually in the repository?

Accepted Answer

A conventional TypeScript layout: src, tests, docs, scripts, package.json, tsconfig.json and package-lock.json, plus knip.json for unused-export checking and skills-lock.json for pinned skills. Skills are wired into three agent configs, .agents/skills, .claude/skills and .pi/skills, alongside a typesafe-ai directory.

Question 6

How mature or maintained is it?

Accepted Answer

Not at all, on the evidence available: a single commit, zero stars, zero forks, zero issues, no releases and no tags. That means no stability guarantee, no changelog and no outside user who has already hit the failure you might hit, so treat it as a reference rather than a pinned dependency.

Question 7

How do I find out what it actually does?

Accepted Answer

Only by reading the code. With no README, the tool names beyond noul, choice and score, their input schemas, return shapes and error behavior are all unstated, and the contents of src, tests and docs are not shown in the listing. Budget time to read src before wiring it into an agent.

TypeSafe Demo MCP

Highlights

Questions about this entry

More Jev resources like this

From the Jev community