MCP for video agents
MCP should expose safe video resources, not uncontrolled camera authority.
Ayneye's MCP direction is a governed interface over durable video artifacts: World-State, timeline, evidence, detector reports, and cost. The public beta uses REST and dashboard first; MCP-style tools remain preview-only until separately audited.
MCP operating model
MCP is useful when an AI tool needs a consistent way to discover resources, call bounded tools, and cite what it used. For video, that interface must be narrower than raw camera control. Ayneye treats MCP as a bridge from an agent runtime to already-materialized artifacts. The server does not give the model the right to interpret identity, infer intent, unlock doors, trigger dispatch, or perform physical actions.
| Layer | What it exposes | What it does not expose |
|---|---|---|
| Resources | scene_graph, timeline, evidence_bundle, detector_report, cost artifacts | Raw unrestricted camera control |
| Tools | read_scene, get_evidence, ask_with_evidence, get_cost_context | Destructive or physical actions |
| Policies | risk class, review_required, allowed claim types, blocked operations | Silent escalation |
| Audit | tool call trace, evidence refs, limits, cost context | Opaque agent behavior |
Choose the right MCP path
Resource catalog
Understand the World-State, evidence, timeline, detector, and cost resources exposed to agents.
Open ->MCPServer architecture
See a preview-safe MCP server design that wraps Ayneye artifacts without granting action authority.
Open ->MCPSafety policy
Review the policy matrix for identity, security, physical access, emergency response, and human review.
Open ->MCPMCP quickstart
Use REST today, then wrap artifact reads in an MCP-style tool boundary for evaluation.
Open ->MCPPolicy contract
Define allowed claims, blocked actions, review triggers, and evidence requirements.
Open ->MCPAgent video tools
Connect MCP-style resources to agent frameworks while keeping the tool surface read-only.
Open ->Current beta boundary
Real signup, dashboard, REST APIs, and artifacts are working evaluation surfaces. MCP is positioned as a preview integration model. Paid checkout, automated email campaigns, CRM writes, and production MCP authority remain disabled until separately enabled and verified.
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.