jhisse

Spoor — AI skill for Claude Code

AI community

Local-first LLM observability.

How to install Spoor

This entry records only its repository, not the path inside it, so there is no exact command to give. Open jhisse/spoor and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Spoor does

Local-first LLM observability. Inspect agent traces, token usage, latency and estimated costs with one binary and one SQLite file.

Alternatives in AI

  • Xerj — Local-first search engine and MCP server for AI agents 12.5k ★
  • Codeburn — Free, local tool to track AI coding token usage and cost across 37 tools and agents (Claude Code, Cursor, Code 9.7k ★
  • Manifest — Real-time cost observability for OpenClaw agents — track tokens, costs, messages, and model usage 4.1k ★

README

spoor: a footprint

spoor

**See where your LLM and agent spend went.**
One binary, one SQLite file, OpenTelemetry in. No account, no key, no cloud.

[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE) [![Go 1.27+](https://img.shields.io/badge/go-1.27%2B-00ADD8.svg)](go.mod) [![No cgo](https://img.shields.io/badge/cgo-none-lightgrey.svg)](#one-small-thing) [![Release](https://img.shields.io/github/v/release/jhisse/spoor)](https://github.com/jhisse/spoor/releases)

[Try it](#try-it) · [Send traces](#send-traces) · [What you see](#what-you-see) · [Connect an agent](#connect-an-agent) · [Operating](#operating) · [Limits](#limits)


A walk through spoor on its demo data: the trace list with what was asked, traces over time, the heat map by cost, the traces worth a look, a search with its hits, one Claude Code trace with the context of every model call, the arithmetic behind a cost, a session turn by turn, a transcript of a tool loop, the blind spots page and the help page

spoor receives OpenTelemetry traces from LLM applications and coding agents, stores them, and shows where the time, the tokens and the money went. Your agent can read the traces too.

What a bill cannot tell you

Question Where spoor answers it
Which call cost the money? Every model call shows the arithmetic: tokens of each type times the price per million, and the price table row used.
Why did the prompt cache stop applying? A call shows only the messages that are new since the previous one. When an earlier message was rewritten, spoor points at the character where it changed and shows the old and new text.
Where did the context grow, and where was the cache lost? One bar per model call s