LangSmith vs TensorZero
LangSmith
langchain.com
TensorZero
tensorzero.comSide by side
| Figure | LangSmith | TensorZero |
|---|---|---|
| Price from | $39/mo | Free |
| Free plan | Yes | Yes |
| Free trial | — | — |
| Open source | — | — |
| Web | Yes | Yes |
| Windows | Yes | — |
| macOS | Yes | — |
| Linux | Yes | Yes |
| API | Yes | Yes |
| Self-hosted | Yes | Yes |
| Plans published | 3 | 1 |
| Spec score · LLM Gateway Software | 9.2 (#1) | 8.0 (#13) |
| Spec score · LLM gateways | 9.3 (#2) | 8.6 (#7) |
| Deployment | Deployment provides managed or self-hosted agent servers with streaming, cron scheduling, authentication, MCP support and more than 30 API endpoints. | The gateway can be deployed with Docker, Docker Compose, Kubernetes using a reference Helm chart, or built from source with Cargo. |
| Integrations | Documented integrations cover OpenAI, Anthropic, Bedrock, Gemini, Mistral, LiteLLM, OpenTelemetry, CrewAI, Vercel AI SDK, PydanticAI, AutoGen and many other frameworks and providers. | Documented providers include Anthropic, AWS Bedrock, Azure, Google, OpenAI, OpenRouter, Together AI, vLLM, and OpenAI-compatible endpoints. |
| Observability | Its observability features include agent tracing, cost tracking, online LLM-as-judge and code evaluations, trajectory monitoring, and webhook or PagerDuty alerts. | TensorZero can store inference and feedback data for observability, optimization, evaluations, and experimentation in Postgres or ClickHouse. |
| Security | LangSmith's trust center lists SOC 2 Type II, ISO 27001:2022, GDPR and HIPAA compliance. | TensorZero's privacy policy says it uses physical, technical, organizational, and administrative security measures to protect personal data. |
| Support | The Developer plan includes community support, and the Enterprise plan lists a support SLA. | The project README directs users to Slack or Discord and offers work teams a Slack or Teams channel by email. |
Plans
In their own words
LangSmith
LangSmith is a commercial agent-engineering platform for building, evaluating, deploying, monitoring, and improving AI agents. It supports offline evaluations on curated datasets and real-time scoring of production interactions. Evaluation methods include human review, heuristic checks, LLM-as-judge evaluators, pairwise comparisons, and custom Python or TypeScript evaluators. Annotation queues route runs to subject-matter experts for feedback that can calibrate evaluations, improve prompts, or augment datasets. The Playground compares prompt versions or model providers, and LangSmith chat can auto-improve prompts. Observability includes agent and retrieval tracing, token and general cost tracking, trajectory monitoring, online evaluations, and webhook or PagerDuty alerts. Traces can arrive through SDKs or an API from LangGraph, custom Python, or other stacks; SDKs are listed for Python, TypeScript, Go, and Java. Hosting ranges from cloud to hybrid and self-hosted Enterprise options. The Developer plan costs nothing, Plus costs $39 a month, and Enterprise pricing is custom.
TensorZero
TensorZero is a free, open-source stack that brings together an LLM gateway, observability, optimization, evaluation, and experimentation. Its gateway provides a unified API for model providers and supports routing, retries, fallbacks, prompt templates, schemas, multi-step workflows, and traffic routing for A/B tests. Documented integrations include Anthropic, AWS Bedrock, Azure, Google, OpenAI, OpenRouter, Together AI, vLLM, xAI, and OpenAI-compatible endpoints. Inference and feedback data go to the database you configure, where they can be queried with SQL; feedback can be tied to individual inferences or episodes. The gateway and UI can run through Docker, Docker Compose, Kubernetes and Helm, or a source build. API key authentication is available for gateway endpoints other than status and health checks. The project is available on GitHub, but is no longer maintained; its repository is archived and read-only. TensorZero is self-hosted and licensed under Apache 2.0. It collects pseudonymous aggregate usage analytics, excluding application data and inference inputs and outputs, with an option to disable collection.