agentfdr

See why your coding agent looped, drifted, or burned tokens

5.0•1 review•

24 followers

See why your coding agent looped, drifted, or burned tokens

5.0•1 review•

24 followers

Visit website

agentfdr reads the transcripts Claude Code already writes and turns them into a flight recording: per-turn timeline, automatic anomaly flags (loops, error streaks, token burns), live watch mode, plan-usage tracking, and cost estimates. Zero instrumentation, 100% local, MIT.

agentfdr Reviews

Name: agentfdr
Rating: 5.0 (1 reviews)

The community submitted 1 review to tell us what they like about agentfdr, what agentfdr can do better, and more.

5.0

Based on 1 review

Review agentfdr?

Reviews

Most Informative

Founder Reviews

No reviews yetBe the first to leave a founder review for agentfdr

Other Reviews (1)

•39 reviews

What's great

Engineering an out-of-the-box, zero-instrumentation debugger that parses the local transaction logs and session transcripts natively generated by Claude Code is a massive workflow optimization for agentic engineering. Developers utilizing terminal-based coding agents frequently experience a "black box" phenomenon where an agent silently burns through millions of context tokens or enters a runaway tool-calling spiral without throwing immediate terminal alerts. agentfdr resolves this friction by introducing an analytics layer that treats agent history like an aircraft flight log. The tool's mechanical highlight is its ability to break down prompt token distribution along a clean chronological timeline—letting you isolate exactly how cache reads, cache writes, and new token inputs inflate across sequential execution frames. Automatically flagging tool loops, error streaks, and cache thrashing using normalized n-gram signatures provides immediate diagnostic clarity, saving developers substantial API capital and token quotas. Running entirely client-side and local-first under an MIT license ensures total workspace security with zero network data leakage.

What needs improvement

The primary structural constraint is that agentfdr is tightly coupled to the specific log formats and transcript generation schemas output by Claude Code. If Anthropic introduces a major update to its terminal tool architecture or refactors its session logging layout, the parsing layers inside agentfdr risk breaking entirely until an open-source patch is committed. To mature as a true cross-platform engine, the telemetry parser needs to scale to support independent frameworks or custom tool-calling runner SDKs. Additionally, while the programmatic CI assertion gate (agentfdr assert --no-loops) is highly efficient at failing out broken agent runs, the application remains primarily a reactive, post-mortem diagnostic desk. Integrating live, hot-swappable context interventions—where a developer can dynamically prune an agent’s inflated history or clip an execution loop mid-session—would elevate it from a monitoring interface into a proactive agent container.

vs Alternatives

I’ve previously evaluated agent telemetry frameworks ranging from enterprise cloud-based LLM observability stacks (like Langfuse, Phoenix, or Arize) to manual text scraping scripts that grep raw local terminal files. Cloud observability platforms are incredibly powerful for tracking production traffic, but they introduce heavy external network piping, complex SDK instrumentation overhead, and data hosting liabilities that make them poorly suited for rapid local debugging. Conversely, custom local logging scripts are notoriously brittle, lack unified cost projections, and fail to calculate multi-turn token distribution models. I chose to leverage agentfdr because it isolates terminal diagnostics into a lightweight, zero-configuration canvas—delivering heavy-duty visibility into token burn profiles and recursive loop anomalies without forcing you to change a single line of your existing project setup.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

5 views1mo ago