The cross-harness, self-improving memory layer for AI agents. Join our community: https://discord.com/invite/zdNChS2fBu
Go agent-ai-cliagent-harnessesagent-securitybrowser-extensionciclaude-codecloudcodexcursordetection-engineeringjevknowledge-basememorymobile-device-managementobservabilitysecuritysecurity-information-and-event-managementsecurity-toolstelemetrytraces
reorder architecture
Justin D'Souza73d94f6c2cd1
diff --git a/README.md b/README.mdindex fdd23aad..f9625e2f 100644--- a/README.md+++ b/README.md@@ -45,6 +45,22 @@ emit agent harness telemetry logs toLearn more in the [Agent Beacon Documentation](https://docs.asymptotelabs.ai/cli).+## High-Level Architecture++Beacon keeps collection, processing, and inspection local to the endpoint while+leaving forwarding under customer control.++<p align="center">+ <img src="images/beacon-architecture.png" alt="Beacon endpoint architecture" width="860">+</p>++- **Agent runtime layer:** Local hooks and OpenTelemetry sources capture+ supported activity from AI agent harnesses on the endpoint.+- **Beacon endpoint layer:** Local processing normalizes events, applies+ retention and redaction settings, and writes durable endpoint telemetry.+- **Output layer:** Teams inspect events in the local dashboard, retain JSONL,+ or forward records into all the major enterprise-grade SIEMs.+## Supported SurfacesBeacon captures supported agent harness activity locally and writes normalized@@ -95,37 +111,6 @@ through standard MDM workflows.| Jamf Pro | macOS package, policy scripts, validation, and Extension Attributes || Fleet | macOS package and user-context deployment helpers |-## High-Level Architecture--Beacon keeps collection, processing, and inspection local to the endpoint while-leaving forwarding under customer control.--<p align="center">- <img src="images/beacon-architecture.png" alt="Beacon endpoint architecture" width="860">-</p>--- **Agent runtime layer:** Local hooks and OpenTelemetry sources capture- supported activity from AI agent harnesses on the endpoint.-- **Beacon endpoint layer:** Local processing normalizes events, applies- retention and redaction settings, and writes durable endpoint telemetry.-- **Output layer:** Teams inspect events in the local dashboard, retain JSONL,- or forward records into all the major enterprise-grade SIEMs.--Beacon filters generic process and runtime metrics, such as Node.js event loop,-V8 heap, process CPU, and process memory telemetry, out of the local endpoint-JSONL by default so agent prompts, tools, approvals, and file activity remain-easy to inspect. Advanced deployments can opt back into those low-level OTLP-metrics with `beacon endpoint install --include-runtime-metrics` or-`beacon endpoint repair --include-runtime-metrics`.--Codex CLI can also emit high-volume internal spans for startup, dispatch, model-requests, streaming, and transport activity. Beacon keeps the default Codex path-low-noise by recording semantic Codex log events for sessions, prompts,-approvals, and tool results while dropping Codex spans. For troubleshooting-Codex OTLP internals, opt back into those spans with-`beacon endpoint install --include-codex-spans` or-`beacon endpoint repair --include-codex-spans`.-## DashboardBeacon includes a local, read-only dashboard for validating endpoint activity