Keep a local, tamper-evident log of everything an AI coding agent does in your terminal.
Audit what tool calls a Claude Code or OpenCode agent actually made during a session.
Replay a recorded coding agent session as a self-contained HTML timeline.
Track subprocesses and network connections an AI agent's process tree opens.
| thisisnithin/snitchit | abc3dz/mixxx | abyo-software/ferro-stash | |
|---|---|---|---|
| Stars | 1 | 1 | 1 |
| Language | Rust | Rust | Rust |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 2/5 | 4/5 |
| Audience | developer | general | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Basic recording needs no privileges, the deepest kernel-level tier requires eBPF capabilities on Linux or Developer Mode and root on macOS.
snitchit is a Rust tool that watches what an AI coding agent does in your terminal and keeps a private, local record of it. You run your agent, like Claude Code or OpenCode, through snitchit, and it writes down what you typed, what appeared on screen, and, once set up more fully, the individual actions the agent itself takes, such as reading or writing files. The project makes three main promises. First, it never sends anything over the network, so nothing you record ever leaves your computer. Second, it only watches, it never blocks or changes what the agent does, and if recording itself fails for any reason the agent still runs normally. Third, the record it keeps is tamper evident: each entry is linked to the one before it using a cryptographic hash chain, so a command called verify can detect if someone edited the log afterward, though the author is clear this protection is local and unsigned, not a full security guarantee. snitchit can watch at three different levels of detail. The simplest level just watches the terminal itself, what you typed and what showed on screen, and needs no special permissions. A deeper level hooks directly into supported agents, Claude Code and OpenCode, to see their internal tool calls. The deepest level watches at the operating system level to catch other programs the agent launches and network connections it opens, using eBPF on Linux or Apple's Endpoint Security framework on macOS, and this level does require extra one time permission setup. To install it you build the project with Cargo and either run your agent explicitly through the snitchit command, or run an install step that wires it into your shell and into Claude Code or OpenCode automatically. You can also generate a self contained HTML file that lets you replay a recorded session in a browser. The project is written entirely in Rust, has one star so far, and its README does not mention a license.
A local, observe-only Rust tool that records everything an AI coding agent does in your terminal into a tamper-evident log, never sending data over the network.
Mainly Rust. The stack also includes Rust, eBPF, Cargo.
No license is mentioned in the README, so reuse terms are unclear.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.