Public agent interface

Give the agent real fire evidence.

A public, read-only MCP endpoint for ChatGPT, Claude, and other compatible clients. Agents can query Lossline’s observed activity, weekly history, and terminology without receiving exact addresses or owner records.

EndpointMCPStreamable HTTP
AuthenticationNonepublic read-only tools
Private lead dataExcludedby database boundary

Connect

One public endpoint.

Add this remote MCP server URL to a compatible client. The server supports current stateless MCP discovery and the established initialization flow for broader client compatibility.

https://api.lossline.app/mcp

Read-only tools

Four precise ways to ask.

Every response includes structured data, a concise model-readable summary, and the canonical Lossline source URL to cite.

search_fire_activity

Filter observed activity by window, state, incident category, or exact public area slug.

list_weekly_fire_reports

List permanent privacy-safe weekly activity snapshots.

get_weekly_fire_report

Retrieve one report using its Monday UTC week-start date.

explain_fire_alert_term

Get a source-aware definition and interpretation limits for a normalized alert category.

Direct APIs

Agents do not need the MCP layer.

The same evidence remains available as ordinary HTTP resources for crawlers, researchers, integrations, and clients that prefer OpenAPI.

Trust boundary

Useful to agents. Safe for the lead.

The public agent server returns broad geography, categories, times, responding-unit counts, and aggregates. Street addresses, coordinates, owner/contact records, customer identities, and watch counts are excluded upstream.

Dispatch data can be delayed, incomplete, duplicated, or recategorized. It is not proof of property damage, owner interest, insurance coverage, or a signed job. Lossline is not an emergency service.