Ayneye MCP · Public Beta
Give your AI agent persistent eyes.
Ayneye exposes a live remote MCP server for video-aware agents. Agents can submit a video source, wait for bounded processing, inspect derived evidence, and ask evidence-grounded questions without sending raw video or raw frames to the language model.
Remote MCP endpoint
https://api.ayneye.com/mcpAuthenticate with a tenant-scoped Ayneye Bearer API key created from your workspace.
Frozen Public Beta tool set
| Tool | Purpose |
|---|---|
| ayneye.create_video | Create and process a tenant-scoped video from a URL or HLS source. |
| ayneye.upload_video | Use an existing file reference or a browser-mediated local-file upload handoff. |
| ayneye.get_video_status | Read processing and evidence readiness for this run. |
| ayneye.get_video_artifacts | Inspect derived artifacts and evidence readiness for this run. |
| ayneye.ask_video | Ask a question grounded only in evidence materialized for this run. |
| ayneye.start_live | Start bounded live execution for the registered source; linked source activation remains scope-enforced. |
| ayneye.get_live_state | Read progressive live-session state, processed frames, events, and terminal state. |
| ayneye.stop_live | Stop the bounded live session while preserving captured evidence for subsequent Ask. |
Processing contract
| Field | Public Beta contract |
|---|---|
| capture_seconds | 1–60 seconds; default 10 seconds |
| hard maximum | 60 seconds per processing session |
| authorized RTSP camera | Controlled Public Beta live source with bounded progressive evidence |
| live Ask | Use ayneye.ask_video against evidence accumulated by the current or completed live session |
| Ask readiness | Evidence must be materialized before Ask |
| Missing video | video_not_found |
| Evidence not ready | evidence_not_ready |
| Raw media to LLM | No raw video or raw frames |
Evidence and authority boundary
Ayneye establishes evidence from the processed run. Outside or model knowledge must not be represented as Ayneye-established evidence. Run-specific evidence must not be presented as a platform-wide capability statement. MCP does not grant physical-control, identity-decision, emergency-dispatch, or destructive authority.
The purpose of this page is to make the MCP boundary concrete for builders, buyers, and reviewers. A useful MCP integration is not a promise that an agent can watch everything and decide everything. It is a disciplined interface over already-materialized evidence. Each tool call should be tenant-scoped, read-only by default, evidence-aware, cost-aware, and auditable. If the request asks for identity, intent, physical access, emergency response, employment action, or destructive change, the tool returns a review-required state instead of pretending that a video artifact is a final decision. This keeps the product useful for developers while making it credible for security, legal, and operations teams.