Before
- Raw footage is hard to query.
- Every follow-up can trigger another visual pass.
- Agents receive vague summaries without traceable evidence.
- Cost and uncertainty are discovered too late.
CPU-first video AI for developers and agents
Turn video into a reusable World-State layer your products can query.
Ayneye helps teams stop asking AI to re-watch the same footage. A bounded processing pass materializes structured state, evidence, timelines, and cost artifacts so dashboards, APIs, and AI agents can reason over compact data instead of raw pixels.
Real signup activeREST beta pathEvidence-first Ask VideoMCP preview boundary
Homepage operating model
Ayneye is not a generic landing page promise. The product path is concrete: register a source, run a bounded processing window, inspect artifacts, ask evidence-grounded questions, and decide whether the workload deserves more capacity. This homepage now routes every major buyer and developer question into a real area of the site.
| Layer | What it means | Where to go |
|---|---|---|
| Product | How video becomes state, evidence, Ask Video, and cost reports | /product |
| Developers | REST workflows, implementation paths, and artifact handling | /developers |
| API | Endpoint map, lifecycle, response model, errors, and examples | /api |
| Docs | Quickstart, schemas, limits, errors, costs, and agent World-State | /docs |
| Use cases | Video agents, live streams, security review, media ops, developer platforms | /use-cases |
| Agents | World-State memory for LangChain, CrewAI, and MCP-style tools | /agents |
| MCP | Preview resources, server sketch, policy boundary, and safety rules | /mcp |
| Pricing | Video-hour planning, cost.json, and repeated-query economics | /pricing |
Choose your evaluation path
Start
Product tour
Understand the end-to-end video → World-State → evidence → Ask Video → cost workflow.
Open →DevelopersDeveloper path
Build with REST, schemas, artifact IDs, evidence refs, and cost reports.
Open →Use casesEvaluation paths
Pick the right path for agents, streams, security, media, or SaaS products.
Open →SafetyTrust boundaries
Understand what is live, beta, preview, disabled, or review-required.
Open →APIAPI docs
Use the beta API lifecycle: add, execute, fetch artifacts, ask, inspect cost.
Open →PricingPricing model
Plan around processed video hours and repeated artifact queries.
Open →What changes after video becomes World-State?
After
- Video becomes structured objects, events, relationships, and timelines.
- Repeated questions use compact artifacts and evidence spans.
- Agents can read state without controlling cameras or taking actions.
- cost.json, fallback, and review flags are visible before scale.
Artifact outputs
Artifact
scene_graph.json
Objects, events, zones, relationships, timestamps, confidence, and review flags.
Open →Artifactevidence_bundle.json
Traceable evidence spans, frame/time intervals, answer references, and review notes.
Open →Artifactcost.json
Processing duration, detector mode, fallback flags, estimated beta cost, retention notes.
Open →Artifacttimeline.json
Temporal event order and refresh markers that make video easier to inspect.
Open →Artifactdetector_report.json
Detector mode, confidence boundary, fallback, and why review may be required.
Open →ProductAsk Video
A question-answer layer that should cite evidence and expose uncertainty.
Open →Safety boundary
Ayneye is designed around read-only evidence before action. The public beta does not enable door unlocking, emergency dispatch, identity claims, intent classification, discipline workflows, destructive actions, hidden billing, automated outreach, or CRM writes. MCP tools remain preview-only until separately audited. Large GPU video models or human review may still be needed for subtle, cinematic, regulated, or high-risk tasks.
Create your Public Beta workspace
Start with one representative video workflow.
Signup is live and creates a limited tenant workspace, dashboard session, and onboarding checklist. The Public Beta plan is intentionally bounded while we collect developer feedback and validate real workloads.
Real signup activeNo card requiredNo checkoutNo automated emailNo CRM write
Frequently asked before a beta test
Is this a replacement for every video foundation model?
No. Ayneye focuses on CPU-first World-State, evidence, timelines, API workflows, and repeated agent queries. High-semantic or regulated workloads may require GPU escalation or human review.
What should we test first?
Use one representative video, run a bounded capture, inspect scene_graph.json and evidence_bundle.json, ask three grounded questions, and compare cost.json with your expected workload.
Does signup really work?
Yes. Signup creates a limited beta account and dashboard session. Checkout, automated email, and CRM writes remain disabled.
Why does pricing focus on video hours?
Processed video hours are the main unit because Ayneye creates durable artifacts that can support repeated queries. Queries are not claimed to be free; they still use application and text-model compute.