Lunary vs PetalTrace
Lunary
lunary.aiPetalTrace
docs.petallabs.ioSide by side
| Figure | Lunary | PetalTrace |
|---|---|---|
| Price from | $20/mo | Free |
| Free plan | — | Yes |
| Free trial | — | — |
| Open source | Yes | — |
| Web | Yes | Yes |
| Windows | — | Yes |
| macOS | — | Yes |
| Linux | — | Yes |
| API | Yes | Yes |
| Self-hosted | Yes | Yes |
| Plans published | 2 | 1 |
| Spec score · AI Agent Observability Tools | 8.2 (#14) | 8.4 (#11) |
| Deployment | Enterprise customers can deploy Lunary in their own VPC using Kubernetes or Docker. | The repository README documents building PetalTrace from source or downloading a release binary and running its daemon locally. |
| Integrations | The integrations directory lists Anthropic, Flowise, LangChain, OpenRouter, IBM WatsonX, Mistral, Grok, Fireworks AI, Ragas, Gemini, AWS Bedrock, OpenAI, Azure OpenAI, LiteLLM and other sources. | The MCP server lets AI agents query trace history, inspect prompts, analyze costs, compare runs, and trigger replays; the docs include Claude Code configuration. |
| What it does | Lunary is a platform for developers building AI chatbots and other LLM-powered applications. | PetalTrace captures AI workflow execution data, including LLM prompts and completions, tool calls, token use, costs, and timelines. |
Plans
In their own words
PetalTrace
PetalTrace is an open-source observability platform for developers examining AI agent workflows. It captures prompts and completions, tool calls, token use, costs, and execution timelines, with a React web interface plus CLI, HTTP API, and MCP server access. Full prompt capture can include system prompts, message history, tool definitions, and model responses. Users can search prompt and completion text, compare runs for structural, content, and cost differences, and replay runs with changed models or temperatures or in mocked mode. The platform accepts standard OpenTelemetry traces, including from applications that do not use PetalFlow. PetalFlow integration adds graph topology, node inputs and outputs, and replay-capable snapshots. Trace data is documented as stored in SQLite, and the software can be built from source or run from a release binary as a local daemon. Authentication is described as future functionality and disabled by default. The public repository identifies an MIT license; no commercial pricing or usage limits are stated.