CLI appTesting and debugging
agentsight
Traces AI-agent processes, files, network destinations, and model calls with eBPF.
- Stars
- 518
- Forks
- 71
- Updated
- Updated Jul 9, 2026
MIT
Traces AI-agent processes, files, network destinations, and model calls with eBPF.
AgentSight observes AI agents at the system level, correlating process trees, shell commands, file operations, network destinations, prompts, responses, tool intent, and model token data. It works outside closed-source CLIs without an SDK, proxy, or vendor integration, and offers live views, SQLite session reports, and offline pprof profiles. It requires Linux eBPF support and elevated probe privileges, and its capture scope includes potentially sensitive prompts and responses.
Resource types
Use cases
Platforms
Create skill evaluations, run benchmarks, and compare performance across models.
Run repeatable evaluations of whether agents discover and use a skill correctly.
Operating systems
Runtime
Capabilities
Audience
Public GitHub facts last synced Jul 10, 2026.
Fetch, merge, and filter Databricks job logs from one CLI command.
Diagnose deployed web apps behind enterprise authentication from the terminal.