tenbin

# Tenbin

A repository for **using [TypeSafe AI](https://docs.typesafe.ai)** (System One API, model Jev) **from a coding agent at design time**. It has three parts.

| Component | Location | Role |
|---|---|---|
| Documentation | [`docs/`](docs/) | Map of the official docs to this repo, workflow diagrams and LLM-integration use cases, design doc, runbook |
| MCP server | [`tenbin/`](tenbin/) | stdio MCP. Question lint, evaluation, batch evaluation (threshold calibration), rerank, taxonomy exploration, cost guard |
| Agent Skill | [`skills/tenbin/`](skills/tenbin/) | Procedure + templates + offline lint for decomposing a judgment into Choice / Score / Noul, measuring, and putting thresholds in code |

Operational procedures (setup, daily work, incident response, maintenance) are in [docs/04-runbook.md](docs/04-runbook.md). Conventions for agents are in [AGENTS.md](AGENTS.md).

## What you can do

Ask your coding agent for a judgment your code needs, and it comes back with a measured design, not a prompt:

| You ask | What you get | Behind it |
|---|---|---|
| `/tenbin` / "Where could Tenbin help this project?" | Proposals grounded in the current project and conversation: evidence, input and judgment type, integration point, uncertainty handling, and the smallest useful trial | skill discovery mode or MCP `tenbin` prompt; no API call |
| "Design and generate Jev code for this project" | Project-specific state/questions, SDK calls, decision logic, failure handling, application wiring and mocked tests | skill contextual integration or MCP `design_integration`; generation needs no API key |
| "Generate state and questions from this context" / "Suggest questions for this input" | Matching `{state, questions}` JSON, field sources, assumptions, type rationale and offline lint;