The model answers.
The gate decides.

An OpenAI-compatible doorway to Navigational Mind governance.

This address isn't a website to read — it's a doorway. Any AI app that already speaks the OpenAI API can be pointed here by changing a single setting, and from that moment every answer it produces is first run through the Navigational Mind gates. The answer still comes back. What comes with it is a verdict and a full record of how that verdict was reached.

See it work → NMA Browser For developers

What a governed answer looks like

Asked: What is the capital of France?

The capital of France is Paris. flagged · hedged

Gate 2 · run on the answer

UNVERIFIED PREMISE: asserted as settled fact, no source — "The capital of France is Paris."

S8 Gatekeeper

Decency PASS · Agency PASS · Navigability PASS

Flagged for the grammar of certainty, not the fact. The gate doesn't doubt Paris — it refuses to vouch for any answer asserted as settled truth without a source, because that is the exact shape a model uses whether it's right or wrong.

What Navigational Mind is

A gate that maps the terrain and leaves the decision with you.

Navigational Mind governance runs an answer through a deterministic pipeline that tests what is being claimed, names what cannot be known from inside the question, and refuses to make a decision that belongs to the person. It doesn't promise truth. It does something narrower and more honest: it tells you the posture of an answer — whether it sourced and hedged, or asserted as if certain — and it shows its working, gate by gate, so nothing is taken on the model's word alone.

How the Gateway works

Four moves, one address.

An app sends a question in the OpenAI format it already uses.
The Gateway forwards it to the Navigational Mind engine, which runs the full gate pipeline.
The governed answer comes back with a verdict and the complete reasoning trace attached.
The app receives a normal OpenAI response — the answer in content, the governance in x_nma.

The three verdicts

Clean

The answer asserted nothing as unearned certainty and gave no direction it shouldn't. Passed through.

Hedged

The answer stated something as settled fact without sourcing or hedging — the form that lets error pass as truth. Flagged.

Held back

The answer made a decision that was the person's to make. Withheld, and the decision returned to them.

The verdict marks an answer's posture and provenance — not whether it is true.

Using it

If your app speaks OpenAI, it already speaks to this.

Point your app's base URL at the Gateway and use your access key. Nothing else changes.

https://gateway.navigationalmind.com/v1

The model field accepts claude, chatgpt, or gemini. A quick test from a terminal:

curl -X POST https://gateway.navigationalmind.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude","messages":[{"role":"user","content":"What is the capital of France?"}]}'

This is an API endpoint — it answers apps, not browser visits. Access requires a key; to request one, get in touch via navigationalmind.com.

Prefer to just watch it think?

NMA Browser lets you ask any question, pick the model, and see every gate fire on the answer — live.

Open NMA Browser →

The model answers.The gate decides.