# Agent Travel API

> Agent-native travel search and validation tools for AI agents.
> Agent Infrastructure Co builds tools for agents themselves; this travel API is the current concrete wedge.
> Best for demos, prototypes, trip-planning copilots, and early travel workflows that need constraint-safe destination/provider search, live evidence where available, provider-ready handoffs, provenance, and bookability state.
> Beta note: current responses expose source/provenance tiers — curated_baseline, model_estimate, live_places, and handoff_required — plus live hotel/place discovery signals; they are not live booking/provider quotes yet. google_aggregator_signals may include AICO-operated Google Flights/Hotels aggregator_fare_snapshots/hotel_rate_snapshots, but these are not live airfare, booking inventory, or provider-backed rates.
> Key trust fields include top-level bookability_status/bookability, match_status, confidence, booking_readiness, provider_handoffs, unsupported_constraints, why_not_bookable_yet, intelligence_basis, live_signals, place_candidates, suggested_itinerary, remote_work_fit, remote_work_snapshot, family_fit, family_snapshot, event_fit, accessibility_snapshot, visa_stay_snapshot, multi_city_snapshot, and event_snapshot.
> Top-level bookability_status is not_bookable for no_match/unsupported_coverage responses and handoff_required for matched beta responses; live_booking_inventory, provider_backed_rates, and live_airfare stay false until provider-backed integrations verify them.
> booking_readiness.inventory_validation_handoff exposes status, next_tool, validates, validation_checklist, and reason so agents can rank here, then validate current_airfare, available_rooms, provider_rates, taxes_and_fees, and availability with live providers before booking.
> provider_handoffs exposes bookability_status, required_external_checks, flight_handoff, hotel_handoff, and place_handoff with provider-ready search params/URLs/checklists; these handoffs are not live airfare, booking inventory, or provider-backed rates.
> Validated primitive proof chain: activation → travel.intent.parse → travel.destinations.search → travel.places.search → travel.plan.validate → travel.provider_handoffs.generate. Branch on match_status, validate candidate_intent_coherence, and only fan out to providers after plan validation passes.
> live-places proof step: travel.places.search now supplies branchable live_places evidence between destination gating and plan validation in the current primitive-composition buyer eval proof.
> Current validated-chain proof: the autumn hiking/food buyer eval selects Point Reyes + Sonoma Coast with candidate_intent_coherence.status coherent, while preserving live_booking_inventory: false, provider_backed_rates: false, and live_flight_fares: false.
> Direct Google benchmark: AICO beats direct Google for agent prefiltering and provider handoff setup, not for provider-backed booking inventory. The eval records baseline Google/browser steps, AICO primitive calls, avoided setup steps, and branchable advantages such as stable JSON, provenance, hard-constraint gates, provider-ready params, and truthful non-booking boundaries.
> Accessibility prompts expose accessibility_snapshot.accessibility_decision_summary, validation_checklist, blocked_reasons, official_validation_required, and recommended_next_tool so agents preserve accessibility as a hard constraint and hand off to live/official validation before final recommendations.
> Multi-city prompts expose multi_city_snapshot.validation_checklist, multi_city_snapshot.blocked_reasons, multi_city_snapshot.recommended_next_tool, and multi_city_snapshot.provider_validation_required so agents use curated_baseline/model_estimate comparison as a shortlist before live rail/hotel/provider validation.
> New hard-constraint fields include user_request, destination_constraints, required_themes, and strict_mode so agents can pass natural language plus explicit region/theme constraints instead of blindly reusing sample JSON; strict city/state geography can also be inferred from natural language when the request says it is strict, says it must be in or must stay in a city/state, or forbids alternatives; match_status reports matched, no_match, and unsupported_coverage so agents do not fill strict requests with off-scope fallbacks. Compatibility alias: shell-only agents may send query instead of user_request; user_request remains canonical and wins when both are present.
> No-match behavior: match_status returns no_match/unsupported_coverage plus hard_constraints, validation_checklist, and live_search_or_provider_search when strict hard geography has no in-scope curated_baseline candidate; agents should not fill with off-scope fallback destinations.
> First-value paid handoff: matched/coherent Travel REST and MCP ranking responses may include commercial_next_step with checkout_source_path /api/v1/travel/search/commercial-next-step, signup/login upgrade URLs, requires_browser_billing_account, limits_delta, source_attribution, and explicit no live booking inventory, provider-backed rates, live flight fares, or booking support truth boundaries.
> API-key paid-intent bridge: shell-only agents can POST /api/v1/commercial/intent with the first-value request_id/source_path to record commercial_intent_requested before browser-authenticated checkout; the response exposes requires_browser_billing_account and paid signup/login URLs without claiming booking support.
> Agent API requests: POST /api/v1/api-requests lets authenticated agents request the next agent-native API they need; if AICO builds an API an agent requests, the requesting account is eligible for $100 of AICO usage credits.
> Agent Commerce API preview: POST /api/v1/goods/search searches and normalizes eBay Browse API provider listing evidence plus craigslist_authorized_snapshot results into used-goods listings, deal_summary, and seller_handoff fields for agents buying/selling used goods.
> Used-goods deal validation: POST /api/v1/goods/deal/validate and hosted MCP goods.deal.validate accept a found listing with title, price_usd, optional listing_url/location/providers, then return comps.summary, price_vs_market, deal_validation, scam_risk_signals, missing_checks, seller_questions, seller_handoff, and truth_boundaries for branchable buyer-agent decisions.
> Used-goods truth boundaries: aico_checkout_supported: false, aico_escrow_supported: false, seller_identity_verified_by_aico: false, and availability_guaranteed: false until AICO has real checkout, escrow, seller verification, and provider availability guarantees.
> Skill Search API preview: POST /api/v1/skills/search accepts an agent's how-to/find-the-best-way query and returns branchable skill/procedure options from skill libraries, including vercel-labs/skills find-skills/SKILL.md and AICO local skill-library entries where relevant.
> Skill Search truth boundaries: executes_skill: false, installs_dependencies: false, guarantees_best_option: false, and live_web_search: false in v0; agents must inspect the selected SKILL.md/provenance before execution.
> Quickstart auth: set AICO_TRAVEL_KEY to a dashboard key and send the bearer header using the segmented shell pattern from llms-full.txt or /docs.

## Canonical URLs
- Home: https://agentinfrastructureco.com/
- Agent Infrastructure Co developer resources: https://agentinfrastructureco.com/developers
- Product: https://agentinfrastructureco.com/travel
- Agent Infrastructure Co API docs: https://agentinfrastructureco.com/docs
- Agent Infrastructure Co OpenAPI spec: https://agentinfrastructureco.com/openapi.json
- Well-known OpenAPI alias: https://agentinfrastructureco.com/.well-known/openapi.json
- Agent Infrastructure Co MCP server: https://agentinfrastructureco.com/mcp
- Well-known agent card: https://agentinfrastructureco.com/.well-known/agent-card.json
- Well-known MCP server card: https://agentinfrastructureco.com/.well-known/mcp/server-card.json
- Markdown home: https://agentinfrastructureco.com/index.md
- Markdown docs: https://agentinfrastructureco.com/docs.md
- Markdown pricing: https://agentinfrastructureco.com/pricing.md
- Well-known markdown pricing: https://agentinfrastructureco.com/.well-known/pricing.md
- Structured data: https://agentinfrastructureco.com/schema.json
- Hosted MCP: https://agentinfrastructureco.com/mcp
- CLI: https://agentinfrastructureco.com/cli
- Signup: https://agentinfrastructureco.com/signup
- Login: https://agentinfrastructureco.com/login

## Direct name lookup
- Query target: agentinfrastructureco developer resources.
- Agent Infrastructure Co developer resources are canonicalized at /developers, /docs, /openapi.json, /.well-known/openapi.json, /mcp, /llms.txt, /llms-full.txt, /index.md, /docs.md, and /pricing.md.
- Search phrases this site explicitly serves: Agent Infrastructure Co API docs, Agent Infrastructure Co OpenAPI spec, Agent Infrastructure Co auth docs, Agent Infrastructure Co MCP server, Agent Infrastructure Co developer resources.

## Self-serve conversion path
- Machine-completable path: POST /api/v1/activation creates a free sandbox account and returns the first bearer API key exactly once; no browser signup, mailbox, or dashboard session is required for sandbox evaluation.
- Browser path: create a free sandbox account and generate a bearer API key from the dashboard.
- Run GET /api/v1/account/me to verify the key and quota state.
- Run POST /api/v1/travel/search and inspect whether the ranked response is useful in the agent workflow.
- Use the response feedback_prompt or POST /api/v1/feedback to submit structured feedback when the ranking is useful, blocked, missing live data, unclear on provenance, or missing a provider handoff; hosted MCP exposes the same feedback path as the feedback.submit tool.
- Request missing primitives with POST /api/v1/api-requests; if AICO builds an API an agent requests, the requesting account is eligible for $100 of AICO usage credits.
- If commercial_next_step is present and the agent wants paid access, POST /api/v1/commercial/intent with request_id and source_path /api/v1/travel/search/commercial-next-step to record commercial_intent_requested from API-key context before opening browser-authenticated billing.
- Shell-only activation keys: 10 requests/hour and 100 requests/month so raw-HTTP evaluation stays low-friction without becoming a production workload path.
- Dashboard-created free keys: 100 requests/hour and 2,500 requests/30 days for browser-capable agents or humans who create an account before first value.
- Hobby: $5/mo flat recurring subscription for higher production-shaped API limits. pricing_model: "flat_recurring_subscription". This is not metered usage billing.
- Recommended paid path: /signup?next=%2Fbilling for new accounts, or /login?next=%2Fbilling for existing accounts; billing is an authenticated upgrade surface, not a public discovery page.

## Public API endpoints
- POST /api/v1/activation — Create a free sandbox account and first API key in one machine-completable request.
- POST /api/v1/travel/search — Rank destinations and return structured trip-planning results.
- POST /api/v1/goods/search — Search and normalize eBay plus authorized Craigslist used-goods evidence with deal validation handoffs.
- POST /api/v1/goods/deal/validate — Validate a found used-goods listing against comps, risk signals, and seller handoff checks.
- POST /api/v1/skills/search — Search skill libraries and return branchable procedure options for agent how-to tasks.
- GET /api/v1/travel/destinations — List supported destinations and their travel metadata.
- GET /api/v1/account/me — Inspect the account and rate-limit state for the current key.
- POST /api/v1/feedback — Submit structured agent feedback tied to a request, blocker, alternative, or missing handoff.
- POST /api/v1/api-requests — Request a missing agent-native API primitive; accepted responses return roadmap_bucket=agent_requested_api and conditional $100 usage-credit eligibility if AICO builds it.
- POST /api/v1/commercial/intent — Record API-key commercial intent from a first-value commercial_next_step response before browser checkout.

## Best fit
- AI travel agents
- itinerary copilots
- hackathon demos
- prototypes that need structured travel ranking without scraping glue

## Why an agent should call this
- Saves tokens and tool calls versus scraping, normalizing, and ranking travel pages from scratch.
- Returns a stable JSON interface with explicit beta warnings and ranked tradeoffs.
- Improves reliability for early travel copilots that need destination ranking, not live checkout.