Developer access
Public playground first, then an instant API key, then the SDK and MCP path into your actual production responses. The docs should feel like part of the product, not a sidecar.
On-ramp
Public proof before private integration.
Use the playground, leaderboard, and findings feed to understand the scoring language before you wire ABIS into your own model outputs.
Authentication
The API key is simple by design.
Public routes stay open. Protected scoring routes accept X-API-Key so teams can prove value fast and automate later.
100
free monthly calls
SDK
recommended path
MCP
desktop integration
Instant free key
Sign up with name, email, and password. Your API key is returned immediately.
Public surfaces first
Leaderboard, findings, and the playground are open so you can evaluate ABIS before integrating.
SDK preferred
Use the SDK for feature extraction and request packaging instead of calling the scoring API by hand.
SDK Quickstart
Use the SDK for the cleanest path into the full scoring pipeline.
pip install abis-sdkfrom abis import ABIS
abis = ABIS(api_key="your-api-key")
result = abis.score(model="gpt-4o", response="Your model response")
print(result.drift_score)
print(result.calls_remaining)from abis import ABIS
abis = ABIS(api_key="your-api-key")
with abis.session(model="gpt-4o") as session:
response = my_ai_call(prompt)
scores = session.score(prompt, response)
print(scores.behavioral_health)
print(scores.entropy)Authentication
The key model is deliberately simple.
Public endpoints like the playground, leaderboard, findings, and model pages do not require authentication. Protected scoring and usage routes use the X-API-Key header.
curl -H "X-API-Key: your-api-key" \
https://abis-drift-monitor-production-b484.up.railway.app/usage/meKey handling
- Your free key is shown once after signup.
- Store it in an environment variable or secrets manager.
- Usage checks are exposed through
/usage/mewhen the backend route is available.
| Plan | API Calls | Notes |
|---|---|---|
| Free | 100 / month | Instant key on signup |
| Pro | 1,000 / month | Adds monitoring and briefings |
| Business | 10,000 / month | Adds operational workflows |
| Enterprise | Custom | Volume pricing available |
Rate limits
Public pages stay open. API volume changes by plan.
The free tier is built to prove the scoring pipeline quickly. Once ABIS becomes part of your daily QA loop, the higher plans open the deeper monitoring and correction surfaces.
Compare plansAPI Reference
Base URL: https://abis-drift-monitor-production-b484.up.railway.app
Get Started
Self-serve endpoints for creating an account and checking your monthly usage.
/auth/signupFreeCreate a free ABIS account and receive an API key instantly.
Request body
{
"name": "Chris",
"email": "you@company.com",
"password": "strong-password"
}Response
{
"status": "registered",
"plan": "free",
"user_id": "you-company-com",
"api_key": "abis_free_...",
"calls_remaining": 100
}/usage/meFreeCheck how many API calls you have used this month.
Response
{
"user_id": "you-company-com",
"plan": "free",
"calls_this_month": 12,
"calls_remaining": 88,
"limit": 100
}Public Surfaces
Open endpoints that power the playground, leaderboard, findings feed, and model pages.
/playground/scoreFreeSurface-only scoring for the public playground. No auth required.
Request body
{
"text": "Paste any AI response here"
}Response
{
"scores": {
"drift": 0.22,
"coherence": 0.74,
"entropy": 0.29,
"complexity": 0.58,
"reasoning_depth": 0.63,
"alignment_stability": 0.81
},
"verdict": "This response shows stable behavioral geometry."
}/public/leaderboardFreeLatest public ranking of monitored models by observed drift.
/public/findingsFreePublished event feed covering drift, anomalies, and version changes.
/public/models/{model}FreeCurrent drift, scorecard snapshot, recent findings, and short trend for one model.
/public/metricsFreeCompact badge-friendly counters for models monitored, version changes, and findings.
Scoring API
Production scoring endpoints. Use the SDK whenever possible instead of constructing requests manually.
/sdk/scoreFreeScore a model response with the full ABIS behavioral pipeline.
The SDK handles feature extraction, signing, and request packaging for you.
Request body
{
"customer_id": "your-customer-id",
"model": "gpt-4o",
"session_id": "session-abc",
"prompt_hash": "sha256...",
"response_hash": "sha256...",
"surface_features": { ... },
"canary_payload": { ... }
}Response
{
"drift_score": 0.23,
"behavioral_health": 0.81,
"entropy": 0.34,
"coherence": 0.89,
"calls_remaining": 87
}Advanced
Operational features for teams running ABIS deeper inside production workflows.
/simulate/correctionBusinessTest a correction strategy against detected drift.
/immune/assessBusinessRun a threat assessment against a behavioral scorecard.
/memory/profile/{model}BusinessInspect longer-term memory and historical behavior for a model.
Ready to score your own responses?
Start free, then move up once ABIS becomes part of your daily model QA loop.