Pricing
Cloud workspaces start at $20 per user per month, and payment is required to activate a workspace. The homepage does not publish usage based tiers, enterprise quotes or separate self-hosted licence terms.
Learn moreLunary
Observability and prompt management for teams shipping LLM apps
Advanced API
What is Lunary?
Lunary is a monitoring, analytics and prompt management layer for LLM applications. It logs prompts, responses and agent traces, tracks cost and user feedback, supports prompt versioning and A/B tests, and can be self-hosted for compliance. Built for engineering and product teams running AI products in production.
Lunary is an observability and prompt management platform for teams building applications on top of large language models. Rather than generating content itself, it sits underneath an AI product and records what happens inside it: every prompt sent, every response returned, the traces that connect them, and the errors that break the chain. The goal is to close the gap between a demo that works and a production system whose behaviour is actually understood. In practice, a developer adds the SDK to an existing codebase (a single line can wrap an OpenAI client, according to the homepage sample) and traffic starts flowing into Lunary's dashboards. From there, teams replay chats, group conversations into detected topics, watch model usage and costs, and read user satisfaction signals such as thumbs up or thumbs down. A playground lets people experiment with prompts and models without writing code, and templates can be versioned and A/B tested so that prompt changes are managed outside the source tree. Human review queues, PII masking, multi-modal support for text, audio and images, and an alerts system round out the operational toolset. The product is aimed at engineering and product teams that run internal knowledge assistants, customer-facing support bots or autonomous agents, from startups to large organisations such as IBM, DHL and Zurich, which the homepage lists as customers. It also suits regulated or European companies: the platform is SOC 2 Type II and ISO 27001 certified, self-hostable via Kubernetes or Docker in a customer's own VPC, and offers role based access control, SSO and masking of personal data. Lunary competes in the LLM evaluation and monitoring space, where the usual comparison points are tracing depth, prompt tooling and data residency. Its distinguishing pitch is the combination of hosted analytics with an open source, self-managed option and a persistent SDK rather than a hosted proxy, which keeps inference in your own account. Buyers evaluating it should confirm which LLM providers and frameworks are covered by the SDKs and how each integration is wired before committing a team.
How do you use Lunary?
1Book a walkthrough or open the interactive demo
Schedule time with the team for a guided session, or step through the interactive product tour on the homepage to see the dashboards before signing up.
2Create a workspace and activate a plan
Sign up for the Team plan from the signup flow. The homepage notes that payment is required to activate the workspace, so have billing details ready.3Connect your LLM app with the SDK
Install the Lunary SDK and monitor your existing client (for example, wrapping an OpenAI client with a single call in Python) so prompts and responses begin streaming in.4Move prompts into templates
Replace hardcoded prompts with Lunary templates so they can be versioned, A/B tested and edited by non-technical teammates without code changes.
5Switch on the tools you need
Enable PII masking, human reviews, alerts and topic classification from the workspace settings, then check the integrations directory for your model provider and data destination.
6Review traces and analytics
Inspect traces and error stack traces, search conversations, and read the analytics dashboards for model costs, user satisfaction and frequent topics before shipping changes.
Pros and cons
Pros
- Covers the full production loop: tracing, chat replays, topic classification, cost analytics and feedback in one dashboardAI
- Prompt templates are versioned and A/B tested outside the codebase, so non-technical teammates can collaborateAI
- Self-hostable with Kubernetes or Docker in your own VPC, backed by SOC 2 Type II and ISO 27001 certificationAI
- Lightweight SDKs wrap existing model clients such as OpenAI without rerouting inference through a proxyAI
- PII masking, RBAC and SSO address the compliance questions that block AI rollouts in regulated industriesAI
Cons
- Pricing is listed only as $20 per user per month with payment required before a workspace activates; volume, usage and enterprise tiers are not publishedAI
- The value depends on instrumenting code first, so non-engineering stakeholders cannot evaluate it without a developer involvedAI
- Feature depth varies by integration, and the homepage groups many data destinations into a single logo wall without detailing what each one actually capturesAI
- Self-hosting shifts operational burden (upgrades, storage, scaling) onto the customer's team and Kubernetes expertiseAI
- Human review and evaluation tooling arrives without published guidance on scoring rigor, so it is not a substitute for a dedicated eval frameworkAI
How much does Lunary cost?
Integrations
SDKs cover providers including OpenAI, Anthropic, Vertex, Azure, Mistral, Llama 2, Replicate and Hugging Face, with framework support such as LangChain. An integrations directory lists data destinations for sending captured LLM data onward.
Learn moreFeatures
Logs and traces for prompts and agent runs, chat replays, error stack traces, instant search, model usage and cost analytics, custom dashboards, topic classification, user feedback tracking, human reviews, prompt templates with versioning and A/B testing, a no-code playground, alerts, PII masking, multi-modal support for text, audio and images, plus RBAC and SSO.
Learn moreFrequently asked questions about Lunary
How much does Lunary cost?
Cloud workspaces start at $20 per user per month, and payment is required to activate a workspace. The homepage does not publish usage based tiers, enterprise quotes or separate self-hosted licence terms.Does Lunary have an API?
Yes, Lunary offers an API.How do you use Lunary?
The walkthrough on this page covers 6 steps: 1. Book a walkthrough or open the interactive demo 2. Create a workspace and activate a plan 3. Connect your LLM app with the SDK 4. Move prompts into templates 5. Switch on the tools you need 6. Review traces and analytics.What platforms does Lunary support?
Lunary is available on Web App.What does Lunary integrate with?
SDKs cover providers including OpenAI, Anthropic, Vertex, Azure, Mistral, Llama 2, Replicate and Hugging Face, with framework support such as LangChain. An integrations directory lists data destinations for sending captured LLM data onward.What are the limitations of Lunary?
Pricing is listed only as $20 per user per month with payment required before a workspace activates; volume, usage and enterprise tiers are not published. The value depends on instrumenting code first, so non-engineering stakeholders cannot evaluate it without a developer involved. Feature depth varies by integration, and the homepage groups many data destinations into a single logo wall without detailing what each one actually captures.
Status
StatusActive
Views0
Outbound clicks0
Added10/22/2026
Platforms
Web App
Pricing
PaidSubscription
