How it works
You give Hailera one API endpoint. Hailera never touches your database and never copies your data: it calls your endpoint live, once per question, with a secret only it has.
Knows in advance
Your schema: resource names, one-line descriptions, field names. Read when you press Test.
Never has
Your database, your credentials, your full data. No copies, no sync.
Keeps
A log line per call: which resource, search or get, result count, time. Not the question.
A customer asks
in ChatGPT, Claude or your website chat
Is my roof repair at 14 Grove St still on?
Hailera's agent decides to look it up
It knows 'jobs' exists from your schema, so it picks search
Hailera calls your endpoint
server to server — the AI model never sees your address or secret
Request to your endpoint
POST https://api.summitroofer.com/map Authorization: Bearer mds_live_7Hq2… {"op": "search", "resource": "jobs", "query": "14 Grove St", "limit": 5}Your API queries your own database
however you like — SQL, your job tracker's API, a sheet
Your endpoint's reply
{"items": [{"id": "job_1182", "address": "14 Grove St, Jersey City", "status": "scheduled", "scheduled_for": "2026-10-01T08:00", "crew": "B"}]}The agent answers from what came back
only your declared fields reach the model; nothing is stored
Yes — your roof job is scheduled for Thursday, Oct 1 at 8:00 AM. Crew B will be there.
