LangWatch

What it does

LangWatch ingests traces of LLM and agent activity, including model calls, tool calls, token use and cost, along with test scenarios, datasets and prompts. It produces searchable traces, simulated multi-turn conversations graded by a judge agent, evaluation scores and pass/fail results that can gate CI. A gateway can also route calls with budgets per person and per team.

Teams write scenarios in plain language, run simulated users in text or voice against an agent, and block pull requests on eval results. The judge reads the full trace so each verdict comes with reasoning. Red-teaming simulations probe for jailbreaks, and OpenTelemetry-based instrumentation works across frameworks. Developers start with the Python or TypeScript SDK; the web app at app.langwatch.ai provides the interface.

It suits AI engineers and platform teams shipping agents to production, and enterprises needing SSO, RBAC and audit logs. The Developer plan is free with 50k events per month and 14-day data access. Growth is €29 per core-seat per month plus usage. Self-hosting runs on Docker Compose or Kubernetes with your own ClickHouse. Event overages and storage beyond 30 days are billed separately.

What it can do

• Simulate multi-turn text and voice conversations with a simulated user against an agent
• Write test scenarios in plain language from Claude Code and run them locally or in CI
• Trace LLM calls, tool calls, tokens and cost via OpenTelemetry and Python/TypeScript SDKs
• Run offline and online evaluations with LLM-as-a-judge, built-in evaluators and CI gating
• Red-team agents with adversarial simulations for jailbreaks and unsafe tool calls
• Route model calls through a gateway with per-person and per-team budgets

Works with

• LangChain
• LangGraph
• CrewAI
• Pydantic AI
• Vercel AI SDK
• Mastra
• Claude Code
• Codex
• Copilot
• opencode

Before you choose

• Billing is usage-based: events beyond 200k per month cost €5 per 100k on Growth, and storage beyond 30-day retention costs €3 per GB.
• Enterprise security features (SSO, RBAC, audit logs, data residency, custom retention) are listed under the Enterprise plan, not the free tier.
• Self-hosting requires your own ClickHouse and infrastructure; the vendor page describes this as keeping data in your environment.

Listed under

• Coding
• DevOps and observability
• Agent testing
• AI gateway
• AI platform teams
• Developers
• LLM evals
• LLM observability

Get started

Developer plan requires no credit card; enterprise evaluation starts with a solutions call.

1. Create an account at app.langwatch.ai (no credit card needed for the Developer plan).
2. Instrument your agent with the Python or TypeScript SDK, or plain OpenTelemetry, to send traces.
3. Write scenarios in Scenario and run them locally or add them to CI; optionally self-host with Docker Compose.

Growth is €29 per core-seat per month plus €5 per 100k events beyond 200k monthly and €3 per GB storage beyond 30 days. Developer is free forever with 50k events per month. Enterprise is custom with invoice or marketplace billing.

Source: langwatch.ai/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Agent Observability & Evals for AI Platform Teams ↗langwatch.ai
• Enterprise AI platform: govern agents at scale | LangWatch ↗langwatch.ai
• LangWatch pricing: free to start, scale by usage ↗langwatch.ai
• About LangWatch | The team building reliable AI ↗langwatch.ai
• Agent Observability & Evals for AI Platform Teams ↗langwatch.ai

Company website

langwatch.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.