Quickstart

Install the CLI

TypeGlish is an npm package. Add it as a dev dependency and run any command with npx typeglish …. Needs Node 20.19+; the Z3 proof tier installs by default (skip with --omit=optional).

$ npm i -D typeglish

Write your first prompt

Near-plain English, with the rules graded by force. Save as hello.tg:

# Role
Agent IS a support agent for Acme, a SaaS analytics company.

# Constraints
- Agent MUST keep every reply to at most 3 sentences.
- Agent ALWAYS escalates a refund request to a human.

Compile it — the checker reads every line as a typed clause:

$ npx typeglish check hello.tg
✓ 1 file — 0 error, 0 warning, 0 info

Break it, on purpose

Add a rule that argues with one you already have, then compile again:

$ echo "- Agent NEVER escalates a refund request to a human." >> hello.tg
$ npx typeglish check hello.tg
hello.tg:6:1  error  logic/contradiction  Logical conflict —
  "escalates a refund request to a human." is both required and
  forbidden. Keep one, or scope the two rules so they cannot both
  apply (IF <condition> THEN …).
✗ 1 file — 2 error, 0 warning, 0 info

That refusal is the product: the contradiction is proven at compile time (by Z3), not left for the model to resolve differently every conversation. npx typeglish --explain logic/contradiction spells out any code.

Fix it, prove it, ship it

Delete the conflicting line (or scope one with IF / UNLESS). Then pin the behavior with a $TEST so the argument stays settled:

$TEST escalation
  - input:: I want a refund — can I talk to someone?
  - expect::
    - offers to escalate the request to a human

$ npx typeglish test hello.tg --dry
✓ hello.tg  coverage: 1/2 rules exercised
✓ 1 prompt — 0 failed
$ npx typeglish build hello.tg
✓ built .typeglish/dist/hello.txt ← hello.tg

--dry validates the suite offline; a live run sends each case to a model (ANTHROPIC_API_KEY). build writes the deployable prompt plus a hash manifest — that artifact, not the .tg, is what your app ships.