Agent Skills as a Memory Layer
Session execution analytics and audit reports for DeepSeek Harness — see how your agent actually worked
Local multi-model fusion gateway for Codex and OpenAI-compatible API relays
A live, session-aware token usage meter for Codex and Claude Code Desktop.
🎼 See what every live Claude Code window is doing — and reply to them. Read-only situational awareness + tmux control, as a CLI, a web cockpit, and an MCP server. Zero deps.
Visualizes LLM agent session logs as step-by-step timelines — user ask, tool call, outcome
Policy-as-code checks for LLM agent tool-call traces
Offline plan drift scoring for LLM agent traces
Trace viewer and human review queue for AI agent sessions — approve, reject, and score outputs
macOS app for OpenAI Codex and Claude Code MCP activity
Claude Code hook that catches the "Fable being quiet" bug — mid-turn messages silently dropped upstream — and has the agent restate them before they're unrecoverable. One file, zero deps, deterministic. Delete when the platform bug is fixed.
Local-first PR review and debugging artifacts for AI agents: trace, replay, eval, scan, dashboard, SARIF, and CI.
Decision-quality observability for LLM agents — scores whether each decision was good, not just how fast it ran. Tool ROI, reasoning coherence, confidence signals. Pure Node.js, zero frameworks.
Trace Lens for agent-built apps — cmd-click any UI surface to see the agent trace behind it. 125-step QA-agent scale fixture, bounded runtime window, and a no-skip Next.js installer e2e that runs a real next build.
Omnodex: Security & Observability for AI Agent Execution
👋 Hi, I'm Saurav — software engineer, AI/ML enthusiast, and open-source contributor based in Seattle
The whole eve stack. On your own machine. Durable Postgres sessions, a Docker sandbox, memory, approvals, schedules, 1,070 tool integrations, and a dashboard that drives the agent.
Local-first observability dashboard for Codex and Claude Code sessions: event stream, conversation replay, token usage, workspace insights, and memory-safe JSONL scanning.
Local-first evaluation, regression testing, and reporting for data agents
A pure-function visualization architecture in which every rendered property is a deterministic projection of append-only operational records — giving provable server/client consistency and byte-exact replay of any past instant, with zero retained gam
Add a description, image, and links to the agent-observability topic page so that developers can more easily learn about it.
To associate your repository with the agent-observability topic, visit your repo's landing page and select "manage topics."