Foglamp

Foglamp

Foglamp is an observability tool by the Vercel AI SDK ecosystem for monitoring AI agents. It provides costs, latency, tokens, distributed traces, evals, and alerts for every generateText and streamTex

AI monitoringobservabilityVercel AI SDKagent tracingLLM analytics
Open site
image for FoglampOpen site

What is Foglamp?

Foglamp is an observability tool built for the Vercel AI SDK ecosystem that monitors AI agents. It captures cost, latency, and quality data for every LLM call, specifically for generateText and streamText functions. The tool provides distributed traces, evals, alerts, and per-agent spend with just two lines of code instrumentation. Its core purpose is to help developers catch bad outputs before end users notice them.

Application scenarios

  • Cost regression detection

    Spot unexpected cost spikes (e.g., a 10× increase) within days of shipping an agent.

  • Quality monitoring

    Track eval scores and catch degrading answer quality over time.

  • Support quality assurance

    Identify when an AI assistant gives wrong refund windows or invents tracking links.

  • Per-agent cost tracking

    Break down spending by model, agent, or customer for budget management.

  • Latency optimization

    Monitor per-agent latency and full call flow to identify performance bottlenecks.

  • Production alerting

    Set threshold rules on cost, latency, and error rate to get notified of issues.

Core Features

  • One SDK instrumentation

    Add two lines of code (`import { foglamp } from "foglamp"` and `const fog = foglamp()`) to instrument every `generateText` and `streamText` call.

  • Per-agent spans and spend

    View individual agent latency, cost, and the full call flow, including sub-tasks like search or writing.

  • Evals with code checks and LLM judges

    Score production traffic using automated code checks and LLM-based evaluations, with a sample pass rate of 94%.

  • Distributed traces

    Waterfall every run and see the exact prompt and response for each span.

  • Alerts with threshold rules

    Set custom thresholds on cost, latency, and error rate to trigger alerts.

  • Cost intelligence

    See exact per-call costs broken down by model, agent, and customer (e.g., Claude Opus $323.12, Gemini $43.06, GPT-5.5 mini $85.05).

  • Per-agent visibility

    Get separate spans, latency, and spending data for each agent in the call flow.

Target users

Developers and engineering teams building AI agents with the Vercel AI SDK. Product managers and operations teams who need to monitor LLM costs, latency, and output quality in production. Support teams that rely on AI agents and need to catch incorrect responses before customers complain.

How to use Foglamp?

  1. Sign up or start free at foglamp.dev.
  2. Add the Foglamp SDK to your project with import { foglamp } from "foglamp".
  3. Initialize the client with const fog = foglamp().
  4. The tool automatically instruments every generateText and streamText call.
  5. Access cost, latency, traces, evals, and alerts from the dashboard.

Pricing and free trial

The website offers a "Start free" option, but no specific pricing tiers or free trial details are provided.

Effect review

Foglamp addresses a real pain point for teams shipping AI agents: the invisible cost and quality degradation that can happen after launch. The user testimonial from a co-founder describes catching a 10× cost regression in week one, which suggests the tool delivers immediate value for monitoring. The combination of distributed traces, evals, and per-agent spend gives developers a single pane of glass for agent observability. For teams already using the Vercel AI SDK, this tool likely saves significant debugging time and prevents customer-facing quality issues.

Frequently Asked Questions

What is Foglamp?
Foglamp is an observability tool within the Vercel AI SDK ecosystem that monitors AI agents, tracking costs, latency, tokens, distributed traces, evals, and alerts for generateText and streamText calls.
How does Foglamp help with AI agent monitoring?
It provides detailed insights into each AI agent call, including performance metrics like latency and token usage, cost analysis, distributed tracing, evaluation results, and real-time alerts.
Does Foglamp support distributed tracing?
Yes, Foglamp offers distributed traces to help you understand the flow and performance of AI agent interactions across different components.
Can Foglamp generate alerts?
Yes, Foglamp provides alerting capabilities to notify you of anomalies or issues based on predefined metrics like cost spikes or latency thresholds.
Is Foglamp free to use?
Foglamp is part of the Vercel AI SDK ecosystem; pricing details should be checked on Vercel's official site, as it may offer free tiers or usage-based pricing.
What data does Foglamp track for each AI call?
It tracks costs, latency, token counts, distributed traces, evaluation scores, and triggers alerts for every generateText and streamText invocation.

Foglamp - AI Tool Detail

Foglamp is an observability tool by the Vercel AI SDK ecosystem for monitoring AI agents. It provides costs, latency, tokens, distributed traces, evals, and alerts for every generateText and streamTex

Category：AI Plugin/Extension

Visit Link：https://www.foglamp.dev/

Tags：AI monitoring、observability、Vercel AI SDK、agent tracing、LLM analytics