Use cases
Observe, diagnose, and improve real agent runs.
Inspect model activity, processes, files, network effects, resources, repeated failures, and the evidence used to evaluate later improvements.
Explore use cases →Product
AgentSight combines runtime evidence, a distributed local-first data plane, supported agent integrations, and an optional coordination layer in one product. Start on one machine and keep the same Node protocol as the fleet grows.
Product map
Use cases, architecture, integrations, and comparisons remain focused pages, but they now sit under one Product entry instead of competing as separate primary navigation items.
Use cases
Inspect model activity, processes, files, network effects, resources, repeated failures, and the evidence used to evaluate later improvements.
Explore use cases →Architecture
Keep detailed runtime data authoritative on AgentSight Nodes while Direct access and the Controller coordinate identity, organizations, discovery, connectivity, roles, and authorization.
See the architecture →Integrations
Connect AgentSight to supported coding agents, local CLIs, process families, containers, and existing observability workflows without rebuilding the agent around an SDK.
Browse integrations →Compare
Compare system-boundary runtime evidence with application traces, OpenTelemetry, gateways, hosted LLM observability, and other adjacent approaches.
Compare approaches →Distributed by architecture
AgentSight scales by coordinating customer-controlled Nodes rather than requiring every detailed session to become a second copy in a centralized telemetry warehouse.
Capture and query detailed runtime data on the Node.
Reach browser-visible Nodes by explicit HTTP(S) endpoints without requiring Controller Relay.
Add identity, discovery, roles, capabilities, entitlements, and relay connectivity through Controller when needed.