One platform for the whole travel transaction. Search to settlement.
Aviare books and manages your travel — fully autonomous, no human agents. Tarmac is the travel-tuned LLM that makes it accurate. OTAIP and HAIP are the open source agent frameworks underneath. Real domain expertise, shipped as code.
What we built
Four products. One stack. End-to-end.
Open core at the bottom — OTAIP and HAIP. Travel-tuned LLM in the middle — Tarmac. Autonomous TMC on top — Aviare. Built to fit together. Sold separately.
Aviare
Autonomous AI Travel Management
Telivity's fully autonomous AI travel management product. No human agents. One LLM. Books, manages, rebooks, refunds, reports, and proactively monitors all travel.
- →Chat-first — no booking wizards
- →Proactive disruption monitoring
- →Stripe Issuing virtual cards
- →ERP / expense / HR integrations
Tarmac
Travel LLM · Enterprise
A travel LLM that doesn't just talk about trips — it runs them. Accuracy, orchestration, governance, execution. End to end.
- →Accuracy — real fares, not guesses
- →Orchestration — full booking lifecycle
- →Governance — every action inside policy
- →Execution — books, tickets, pays
OTAIP
23 stars · 8 forks
Open Travel AI Platform. Domain-specific AI agent orchestration for air. Agents across the full transaction — search, price, book, ticket, settle.
- →Adapters: Duffel, Sabre, Amadeus
- →Channels: ChatGPT, Claude (MCP)
- →Pipeline contract gates
- →Apache 2.0
HAIP
11 stars · 7 forks
Open-source, API-first hotel PMS. TypeScript / NestJS / Postgres. Multi-tenant by design. The hotel-side complement to OTAIP.
- →Multi-tenant PMS architecture
- →REST API + OpenAPI auto-gen
- →Built-in AI agents
- →Apache 2.0
Why Telivity
Real domain knowledge, shipped as code.
18 years in airline distribution
GDS/NDC, TMC operations, IRROPS, BSP settlement, fare rules. We've lived the failure modes — the agents encode them.
Open core, not open-washed
OTAIP and HAIP are Apache 2.0. The same code that powers Aviare in production. You can read it, fork it, run it.
Agent-orchestrated, not LLM-glued
Typed agents with contract gates. The LLM handles intent. Specialized agents handle every step. Hallucinations get caught at the schema layer, not in production.
Built to be consumed by AI
Native ChatGPT (OpenAPI) and Claude (MCP) channels. The travel layer for the next wave of consumer assistants.
Talk to us.
Building on OTAIP or HAIP? Evaluating Aviare for your company? Want Tarmac for an enterprise deployment? Drop us a line.