naisslab/agentshark is a tool for examining the execution logs of coding agents such as Claude Code and Codex locally. It provides a native macOS app written in Swift and a web viewer that can be used on multiple operating systems. For now, both have to be built directly from source.
AgentShark indexes execution logs into a local SQLite database and shows the actions in chronological order. Failed commands and success claims that have not yet been verified can be reviewed first, and analysis results can be compared against the original log bytes. The configuration focuses on tracing the cause of failures and the fix process at the time. It also provides a read-only MCP server for looking up past execution logs. The README also introduces sensitive-information masking and offline container replay. It is a tool aimed at people who want to examine logs without sending them to an external service.
However, what can currently be confirmed is limited to Claude Code and Codex session logs, and there is no registered release. Since both the macOS app and the web viewer must be built directly, there is no confirmed distribution you can install and use right away. The MIT license and the local indexing show the project's public terms and how it handles logs, but they do not prove the actual scope of use or the level of completeness. Therefore, for now, it is better regarded as an early-stage tool whose log analysis features can be reviewed by building it directly.