stormzhang/token-tracker — Track token usage across local AI agents (Claude Code, Codex) — CLI dashboard with cost analysis, ra…

Description and use

**Brief Summary (2‑3 sentences)**
stormzhang/token‑tracker is a Python‑based CLI tool that monitors token consumption for local AI agents such as Claude‑Code and Codex. It offers a real‑time dashboard with cost breakdowns, rate‑limit alerts, and session‑level tracking, making it easy to audit and optimise usage across prototype or internal AI workflows. With 78 ★ on GitHub and recent updates, it provides a lightweight observability layer for developers building RAG pipelines, agent orchestrations, or other AI‑driven features.

**Value**

• **Cost visibility** – instantly see how many tokens (and thus dollars) each request consumes, helping teams stay within budgets.
• **Rate‑limit safety** – proactive warnings prevent throttling errors that can stall development or production jobs.
• **Session tracing** – per‑session logs let engineers debug prompt engineering, compare model behaviours, and benchmark tooling choices.
• **Low‑friction integration** – works via API/SDK hooks or a simple CLI, so you can add token observability without rewriting existing model‑calling code.

**Practical Adoption Path**

1. **Prototype stage** – add the library to your Python environment (pip install token-tracker) and wrap your existing model‑call functions with the provided decorator or CLI wrapper.
2. **Validate** – run a few test prompts; the dashboard will display token counts, cost estimates, and any rate‑limit warnings.
3. **Iterate** – use the data to optimise prompts, choose cheaper models, or batch requests.
4. **Internal rollout** – embed the tracker in CI pipelines or internal notebooks to enforce cost policies across teams.
5. **Production hardening** – if needed, export the CSV/JSON logs to a central observability platform (e.g., Prometheus, Datadog) and configure alerts.

**Production Readiness**

• **Maturity**: Medium. The project is actively maintained (last commit 2026‑05‑12) and has modest community traction (78 ★, 9 forks).
• **Dependencies**: Pure Python with standard HTTP/CLI libraries, making integration straightforward.
• **Risks**: License and long‑term maintainer commitment still need verification; security review of the token‑exposure path is recommended.
• **Recommendation**: Suitable for prototypes, internal tooling, or as a monitoring adjunct in production after a brief security and licensing audit, and after adding redundancy (e.g., fallback logging) for mission‑critical workloads.

Problem

stormzhang/token-tracker helps add AI capability without starting from a blank model stack.

Outcome

Medium: useful for prototypes or internal workflows, with dependency and maintenance checks before production.

Workflow

• prototype AI features
• build RAG or agent workflows
• evaluate model tooling

Evidence

Track token usage across local AI agents (Claude Code, Codex) — CLI dashboard with cost analysis, rate limit monitoring, and session tracking

Source