Developer portal

Your AI has no integrity layer.
One API call fixes that.

DreamerOS is not another AI wrapper. It is a checked execution pipeline. Every response passes through a question shaper, integrity checks, drift detection, and a follow-up forecaster before it reaches your application.

Ship AI features that are structured, checked, and auditable from day one.

Start at your depth

An API is a doorway between two pieces of software. Think of it like a restaurant window. Your application sends an order in a known format. DreamerOS does the work behind the window and returns a checked result in a format your application understands. You do not need to see the kitchen to use it. Engineers can inspect every station when they need to.

I do not write code. Use the product, not the doorway. The same checks already run inside the DreamerOS app. Ask a real question there and let the API stay invisible. Open DreamerOS
I am building something. Send one request. Get structured proof back. Start with the Node quickstart or raw REST. Add memory, engine selection, verification, and receipts without rebuilding the checks layer. Open the quickstart
I evaluate systems. Inspect boundaries, failure paths, and evidence. Trace the stages, authentication, tier gates, endpoint contracts, audit exports, and the exact record left behind. Inspect the architecture

The five-line quickstart

// npm install @dreameros/sdk
import { DreamerOSClient } from "@dreameros/sdk";

const client = new DreamerOSClient({ apiKey: "dros_your_key" });

// Store a fact once...
await client.remember({ content: "We ship Tuesdays. Never Fridays." });

// ...recall it from any connected tool
const hits = await client.recall({ query: "when do we ship" });

// Send a question to the right engine
const routed = await client.route({ message: "Draft the launch email" });

// chat() and chatStream() run the full checked chat
// pipeline with your session token - quickstart below.

See checked outputs

What runs on every call?

Architecture

What runs on
every API call.

Six stages. Your user sees none of them. Your team can audit all of them.

The six stages, in order

Input. Your prompt.
Layer 1. Question Shaper.
Layer 2. Smart engine selection.
Layer 3. Integrity check.
Layer 4. Forecaster.
Output. Checked response.

Every stage is logged. Every stage is queryable. Your compliance team will thank you.

Under the hood, in depth

Show me the endpoints

Endpoints

Trust and integrity primitives
behind one Bearer token.

REST API. JSON in, JSON out. Bearer auth. Rate limits scale with your tier. Built on the core pipeline: verify any AI output, underwrite a claim, issue bonds against integrity scores, prove provenance, and export compliance audit trails.

Consumer endpoints (any tier)

POST /api/v1/verify. Submit any text claim - yours or any AI's - for verification. Four depth levels: light is free on every tier, medium is $0.02 per call on every tier and is what you get if you do not name a depth, deep is $0.05 per call (Pro+), exhaustive is $0.25 (Elite). More depth means more detail in the response. Returns a verdict, a score, flags, and an ed25519 signature you can hand to your own customers.

{
  "verdict": "verified",
  "score": 0.94,
  "depth": "deep",
  "flags": [],
  "signature": "ed25519:...",
  "cost_usd": 0.05
}

GET /api/v1/fingerprint/export. Export your behavioral fingerprint as a signed, portable passport - five behavioral axes plus engine trust scores, signed with ed25519. Any downstream AI can verify and import it via MCP, so your tone, preferences, and trust history follow you to every tool you use.

GET /api/v1/leaderboard/caught-this-week. Public, no-auth feed of the top 20 users by high and critical integrity catches in the past 7 days. User IDs are SHA-256 pseudonyms - no PII. Live proof the system catches real drift in real time, not just benchmark cases.

GET /api/v1/dashboard/cross-service-feed. Unified event feed across Sentry, Railway, Vercel, Supabase, and GitHub for your account. One endpoint, one shape, one auth. Tells you what just broke without opening five dashboards. Per-source failures degrade to a warning - the whole feed never goes blank. Elite.

POST /api/v1/intent/clarify. Send an ambiguous prompt, get back the structured version that would actually run. Public for any tier; Pro+ unlocks the full layer - emotion signal, depth hint, and an ambiguity score. Use it to pre-flight a prompt before paying for execution, or to figure out why an AI misread you.

B2B and Enterprise endpoints

POST /api/v1/underwrite. Submit an AI-generated decision or recommendation; get back an underwritten verdict you can stand behind - a fidelity score, a depth audit, and a signed certificate. $0.10 per call, Pro and above. For when "the AI said so" is not a defensible answer.

GET / POST /api/v1/bonds/*. Issue and list bonds backed by integrity scores. GET /listings returns public bond inventory. POST /issue mints a new bond against an underwritten claim. Bonded performance instead of unbacked promises. Pro+ to issue, public to read.

{
  "bond_id": "bond_a1b2c3...",
  "backing_score": 0.96,
  "underwriter_signature": "ed25519:...",
  "face_value_usd": 10000
}

POST / GET /api/v1/provenance/*. Anchor an AI output to a verifiable provenance chain. POST /claim registers a new claim with engine, model, integrity metadata, and a content hash. GET /{claim_id} resolves the chain. C2PA-aligned. Use it to prove which model said what, when, and under which integrity check.

GET /api/v1/audit/export. The full append-only integrity events trail for your account as a compliance-ready bundle - JSONL plus manifest plus signature. One endpoint, no manual assembly, ready for SOC, ISO, and internal review. Elite only.

POST /api/v1/ifp/score. Integrity as a service: score any prompt and response pair against the Intent Fidelity Protocol. Returns the fidelity score and what failed when something does. In allowlist beta today - request access and wire it into your own pipeline.

MCP dreameros_slop_check. Anti-Slop Check. Catches weak AI output patterns - defensive guards, narration comments, premature abstractions, half-finished stubs - before they reach your users. Pro tier MCP tool, advisory on all tiers.

Pricing. Usage-billed endpoints land through integrity_events.cost_usd - no separate metered-billing setup. Costs land on your Stripe invoice at month-end, itemized by endpoint. Underwriter is $0.10 per call, Bonds is per-bond face-value percentage, Provenance is included with Pro+, Audit Export is included with Elite. No surprise charges - every call is logged before it bills.

Integrity checks inside any AI tool (MCP)

DreamerOS ships as an MCP connector for any compatible AI client. Plug integrity checks directly into Claude, ChatGPT, Gemini, Grok, Perplexity, or any MCP-compatible tool.

Light connector (free). Behavioral integrity. A contract of more than 30 enforced principles shapes output quality without computational transformation. Works with all supported engines.
Pro connector ($39.99/month). The full pipeline. Question shaping, integrity validation, smart engine selection, token efficiency tracking, and connected memory. The complete integrity layer running inside whatever AI tool you already use.
Integration. The connector endpoint is live at the DreamerOS gateway. Authentication is via API key. Tier-based engine selection runs server-side - the connector surface looks the same on Light and Pro, but the pipeline depth changes with your subscription.

See the Connect page for client setup

The live receipt ledger

How do I start?

Get started

Three lines.
Checked output.

Node SDK today. Python SDK coming soon. Raw cURL always works.

Pick your stack

Node.js. Typed client for chat, streaming, memory, and engine selection. Works with Next.js, Express, and serverless. npm install @dreameros/sdk
Python. Async-ready. Type-hinted. Works with FastAPI, Django, and Flask. Python SDK - coming soon.
cURL / REST. Standard REST. JSON in, JSON out. Bearer auth. Drop it into any stack. Authorization: Bearer dros_...

import { DreamerOSClient } from "@dreameros/sdk";

// chat auth: your session token. remember / recall / route: your dros_* key.
const client = new DreamerOSClient({ apiKey: token });

const response = await client.chat({
  message: "Should I open a second location?"
});

console.log(response.content);
// "Before expanding, audit unit economics on location one.
//  Most second-location failures are location-one problems
//  that got exported."

console.log(response.engine, response.model);
// which engine answered, and under which model

curl -X POST https://dreameros-scs-gateway-production.up.railway.app/api/v1/verify \
  -H "Authorization: Bearer dros_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Churn dropped 40% after the pricing change",
    "depth": "light"
  }'

# Every call returns a signed verdict and
# lands in your append-only audit trail.

Integrity checks that ship with your code

CI/CD ready. Drift detection in your pipeline. File hash verification on every deploy. Integrity scoring in your tests.

Drift Sentinel. Automated behavioral tests on every deploy. Define expected behaviors in YAML. DreamerOS runs test prompts against the live system and checks outputs against your baselines. Drop below 90% pass rate and the pipeline stops.

# .github/workflows/drift.yml
- name: Run Drift Sentinel
  run: dreameros drift-test --baseline v5.yaml

File Hash Verification. Every integrity file is SHA-256 hashed in a registry. The gateway checks hashes on startup. If any file has been tampered with, the system will not start. Add verify_canon.py to your CI to catch drift before it ships.

# Block deploys on canon mismatch
- name: Verify Canon Integrity
  run: python verify_canon.py

Fail-Closed Gateway. If the integrity check system is unreachable, the API returns 503 - not a degraded response. If file hash verification fails, the gateway will not start. If direction sets fail to load, no requests go through. Integrity is never optional.

# No silent degradation. Ever.
"error": "integrity_unavailable"
"status": 503

What are the boundaries?

Integrity

Every request authenticated,
authorized, and accounted for.

API keys are scoped to your organization and tied to tier permissions. Rate limits scale with your plan. Every request is logged in the accounting layer.

Keys, limits, and the record left behind

Keys are generated from your DreamerOS dashboard. Each key is bound to a single organization with configurable permission scopes. You can rotate, revoke, or scope keys to specific endpoints at any time.

Rate limits are applied per key and vary by tier - from development sandbox to production throughput. Every request is logged for auditability.

Who this is for

Builders who take AI seriously. Whether you are shipping your first AI feature or running AI at organizational scale.

Solo devs and indie hackers. "I need AI features but I don't have a team to build safety infrastructure." Ship AI features with built-in integrity checks from day one. Hallucination detection, integrity checking, and audit trails come with every API call. Focus on your product, not AI reliability plumbing.
Product teams and SaaS platforms. "Our AI features are inconsistent. Users don't trust the output." Embed DreamerOS into your existing product. Every response includes integrity metadata so you can build trust into your UX - not just hope for it.
CS students and AI researchers. "I want to understand intent-verified AI architecture, not just consume APIs." DreamerOS exposes its integrity internals. Every response includes engine selection metadata, integrity scores, and engine selection rationale. Study how checked AI systems work, build research projects around drift detection, or benchmark integrity scoring against your own models.
Enterprise and Elite teams. "We need checked AI across our entire org with role-based access and audit trails." Multi-engine orchestration, organization-wide session memory, role-based API keys, full audit logging, custom domain modules, and webhook-driven automation. The integrity infrastructure you would eventually build yourself - already built and running in production.

Ready to build?

First action

Ready to
build?

Request API access and start shipping AI with integrity checks built in from day one.

Other doors in

Try DreamerOS free DreamerOS GPT

Or see all tiers side by side for the full feature breakdown.

Request API access

Your AI has no integrity layer. One API call fixes that