Overview
LangWatch Claude Code usage tracking is a focused observability workflow for developers and teams running coding agents. It captures sessions from Claude Code, Codex, OpenCode, Copilot, Cursor, and Pi through an OpenTelemetry-style setup, then turns model turns, tool calls, token classes, and theoretical cost into searchable traces.
The product is useful because coding-agent spend is often hard to understand. Claude subscriptions, cached tokens, long sessions, multiple machines, and background tool calls make a simple local log viewer incomplete. LangWatch gives teams a central place to see what happened, how many tokens were used, where costs concentrate, and which prompts or workflows need improvement.
This listing focuses on the Claude Code usage product page, not the entire LangWatch platform. LangWatch also offers broader AI data analysis, LLM observability, evaluation, and governance capabilities for production AI teams.
Key Features
Claude Code trace capture - Run
npx langwatch claudeto configure Claude Code's native export path and send sessions into LangWatch as traces.Multi-agent usage tracking - The official page says the same treatment works across Claude Code, Codex, OpenCode, Copilot, Cursor, and Pi.
Token class visibility - LangWatch separates input, output, cache read, and cache write token classes, which helps explain why long agent sessions can look expensive or cheap depending on caching.
Theoretical bundled-plan cost - For Claude subscription sessions, LangWatch can mark usage as bundled while calculating the theoretical API-list-price cost.
Tool-call and terminal context - Developers can inspect what the agent did, including tool calls and session traces, instead of relying on a rough end-of-session summary.
Dashboards and analytics - Usage, events, storage, and cost by user, prompt, and model are visible in the dashboard, making it relevant to AI data visualization and team governance.
Pricing & Plans
LangWatch has a free Developer plan and paid Growth plan. The pricing page describes Growth as €29 per core-seat per month, with usage charges beyond included allowances.
| Plan | Price | Includes |
|---|---|---|
| Developer | Free | Start sending events, suitable for individual use and trials |
| Growth | €29/core-seat/month | Paid team usage, unlimited evaluations and simulations, additional event/storage billing |
| Enterprise | Custom | SSO, RBAC, self-hosting, marketplace billing, SLA, enterprise controls |
The pricing page states that billing combines seats and usage. Users should confirm current event allowances and storage retention before budgeting production-wide tracking.
Best For
- Engineering teams using Claude Code, Codex, Cursor, OpenCode, Copilot, or Pi across multiple machines
- Managers trying to understand coding-agent usage by user, prompt, model, and workflow
- Developers who need trace history rather than only a local JSONL log
- AI platform teams connecting coding-agent workflows to broader observability practices
- Organizations that want to compare bundled subscription usage with theoretical API costs
FAQ
What does LangWatch Claude Code usage tracking do?
It captures coding-agent sessions as traces, including model turns, tool calls, token classes, and theoretical spend. The result is a searchable usage history rather than a single local log.
How do I start tracking Claude Code usage?
The official page says to run npx langwatch claude. That configures Claude Code's native export path so sessions land in LangWatch.
Does LangWatch work with Codex?
Yes. The official page says Claude Code, Codex, OpenCode, Copilot, Cursor, and Pi can export through the same endpoint and receive the same trace and cost treatment.
What does "bundled" cost mean?
When sessions run under a Claude subscription, the actual billing may be bundled into the subscription. LangWatch can still calculate a theoretical API-list-price cost to show what that usage would have cost through API billing.
Is there a free plan?
Yes. LangWatch offers a free Developer plan. The Growth plan starts at €29 per core-seat per month, with usage and storage components beyond included allowances.
Can LangWatch self-host?
The pricing FAQ says LangWatch can run fully self-hosted with Docker Compose on a user's own ClickHouse setup, with enterprise controls added on the Enterprise plan.
Who should skip LangWatch Claude Code usage tracking?
Skip it if you only run occasional local sessions and do not need multi-machine traces, dashboards, cost analysis, or team-level visibility.
