Multi-model intelligence,
by API.
Run any prompt through GPT, Claude, Gemini, and Grok in parallel. Get a smarter answer than any single model produces, with the deliberation trail to back it up. One API, four models, three intelligence levels.
# Get a multi-model answer for a high-stakes decision curl https://api.truverif.ai/v1/queries \ -H "Authorization: Bearer $TVAI_KEY" \ -H "Content-Type: application/json" \ -d '{ "mode": "justify", "prompt": "Should we approve this $2M credit line for ACME Corp given their last 4 quarters of financials?", "models": { "openai": "gpt-5.4-2026-03-05", "anthropic": "claude-opus-4-7", "google": "gemini-2.5-pro", "xai": "grok-4-1-fast-reasoning" } }'
Three intelligence levels
Pick the right tradeoff for the job
Same API contract, three orchestration strategies. Unify is fast synthesis. Justify is multi-round consensus for high-stakes decisions. Verify is sourced research for when accuracy is non-negotiable. Switch modes per request — credits scale with how much work each mode does.
Unify
fastRun all four models in parallel, then synthesize a single answer. Faster than calling any one model individually and broader in perspective. Use for fast lookups, routine reasoning, content drafts, and anywhere a single LLM call would have worked.
Justify
consensusMulti-round deliberation. The four models see each other's answers and revise their positions. Disagreements surface as structured signal — not buried in prose. Use for high-stakes decisions: financial analysis, legal review, complex reasoning where being wrong is expensive.
Verify
researchFull research mode. Each model investigates the prompt with web access, extracts claims, grounds them in sources, and cross-checks the others. Returns a structured per-claim verdict. Use for fact-checking, compliance review, research summaries, and any workflow where every assertion needs a citation.
1, 2.5, and 5 credits assume Standard-tier models across all four families. Actual credits per query also vary with query complexity and the amount of web research required.
Why multi-model
Why multi-model matters
Single-model AI has predictable failure modes — and they don't show up in benchmarks. Four models, asked the same question, fail in different ways. The disagreements become signal.
Smarter on every task
A single model can be persuasive and wrong. Four models reasoning in parallel surface uncertainty that a confident single answer would have hidden. The aggregated output is genuinely better — not just more cautious.
Audit trail by default
Every response includes the four underlying model answers, the conflicts detected, and how disagreements were resolved. When the decision matters, you can show your work — without building telemetry yourself.
Async by design
Justify and Verify run for 60–300 seconds. The API is built for that: 202 + polling or SSE streaming, idempotency keys, and result endpoints that survive client retries. No 30-second timeouts to design around.
Same credits as web
Your existing TruVerifAI subscription works. API requests deduct from the same credit pool — no separate billing, no second invoice, no new pricing to learn.
Plug in your own
Or bring your own models
Layer in models you already pay for — AWS Bedrock, GCP Vertex AI, OpenAI, Anthropic, Google AI Studio, or any OpenAI-compatible host. Same orchestration, your credentials.
Pass them per request
Send custom_models on POST /v1/queries with the byom_<id> strings GET /v1/models lists in its custom group. The BYOM participates alongside (or instead of) the four defaults.
Your keys, your billing
TruVerifAI treats BYOM inference as zero variable cost — you pay your provider directly, we charge only the mode's orchestration credits.
Cloud-account inference
AWS Bedrock and GCP Vertex AI BYOMs run inference inside your own cloud account, on your IAM credentials or service account.
Use as the brain of your agent
OpenAI-compatible — and smarter
Point any OpenAI-compatible agent framework at TruVerifAI. Change one base URL. Get a brain that runs four models in parallel and only acts on consensus.
from openai import OpenAI # Use TruVerifAI as the brain for high-stakes decisions client = OpenAI( base_url="https://api.truverif.ai/v1", api_key=os.environ["TVAI_KEY"] ) response = client.chat.completions.create( model="truverif-justify", messages=[{"role": "user", "content": "..."}], tools=[get_position, get_market_data, place_order] )
Tools split by intent
Read tools inform the deliberation. Write tools fire only after consensus. Your agent can place orders only when four models agree.
Skills-compatible
Load Anthropic's open Skills format directly. Inject domain expertise into every model in the deliberation — financial analysis, legal methodology, compliance frameworks.
Built for high-stakes loops
Designed for agents where being wrong is expensive: trading, underwriting, contract review, medical decision support. Latency is a feature, not a bug.
Where multi-model intelligence wins
Five workflows we built this for
Five workflows where calling multiple models in parallel produces materially better outcomes than calling one.
Investment and decision agents
Use TruVerifAI as the brain for trading agents, research assistants, and underwriting workflows. Four models deliberate; the action only fires when consensus is reached. Get the audit trail regulators ask for, built in.
Editorial review at scale
Run drafts through Verify before they ship. Catch factual errors, missing context, and unsupported claims before readers do. Structured output queues only the ambiguous cases for human review.
Regulated content review
Financial disclosures, legal updates, healthcare summaries. Justify mode surfaces where models disagree on interpretation — the exact spots that need a human compliance officer's attention, not a rubber stamp.
Research with citations
Earnings calls, filings, analyst notes, scientific literature. Verify mode returns per-claim sources so the summary you ship is traceable to the original document — not the model's training data.
Fact-check pipelines
Plug into your CMS. Automated fact-checks on submission, structured output, escalation queue for cases where models disagree. Keep editorial throughput without sacrificing rigor.
Pricing
Same plans as the web product
API requests draw from the same monthly credit pool. No separate billing, no new pricing to learn.
Free
$0forever
50 credits, lifetime
- All three modes
- 60 RPM rate limit
- Community support
Basic
$12/ month
100 credits / month
- All three modes
- 60 RPM rate limit
- Email support
Pro
$30/ month
300 credits / month
- All three modes
- 60 RPM rate limit
- Priority email support
Enterprise
Custom
Custom credit volume
- Dedicated support channel
- Custom rate limits
- Volume pricing
Talking to a regulated industry buyer? Contact us to discuss procurement, custom rate limits, and volume pricing.
Quickstart
Get started in 60 seconds
No SDK install needed for the first call.
Create an account
Sign up at truverif.ai. 50 free credits land in your account immediately.
Generate a key
Open Settings → API → Create key, accept the AUP. The secret is shown exactly once — copy it.
Run your first query
Send a curl request to /v1/queries. Get back a request_id, then poll or stream until done.
Use it as an agent brain
SoonWire it into LangChain, LangGraph, AutoGen, or your custom agent. We're OpenAI-compatible — change one base_url and you're done.
Build the smartest brain in your stack
Issue a key, read the docs, swap your model provider with one line of code. We’ll handle the rest.