
Phoenix
β Text source-checked Oct 23, 2026Open-source AI observability and evaluation platform for tracing and improving AI agents.
Source-checked against the vendor's site Β· rebuilt from the vendor's own pagesProfile updated Link checked Pricing captured How we rate
In short
Phoenix is an open-source, local-first platform for tracing and evaluating AI agents. It supports OpenTelemetry and runs locally or on Kubernetes.
Phoenix is an open-source, local-first platform for tracing, evaluation, experimentation, and prompt iteration. Phoenix allows users to trace, evaluate, and improve AI agents.
The vendor says that Phoenix traces every step of an agent's execution, including prompts, retrievals, tool calls, and outputs. Phoenix supports native OpenTelemetry, allowing traces to work with existing tools. Users can build evaluations to score outputs and catch issues before they reach end users. The experimentation workflow allows users to create datasets from traces and run experiments to ship improvements.
Phoenix can be run locally on a laptop in under a minute. Phoenix can be deployed and scaled on Kubernetes using Helm. Phoenix is vendor-agnostic and works with any model, framework, or language.
According to Phoenix: Phoenix is an open-source, Local-first platform for tracing, evaluation, experimentation, and prompt iteration. Its pages list this strength: Phoenix is ELv2 licensed and built in the open with community contributions.
β The AI Tool Bible editorial team
Pros
- β Phoenix is ELv2 licensed and built in the open with community contributions.
- β Phoenix uses native OpenTelemetry support to avoid proprietary lock-in.
- β Phoenix is vendor-agnostic, working with any model, framework, or language.
- β Users can self-host Phoenix to keep sensitive data on their own infrastructure.
Pricing (captured )
| Tier | Monthly | Key features |
|---|---|---|
| AX Free | $0 |
|
| AX Pro | $50 per month |
|
| AX Enterprise | Custom |
|
Prices change; we show the grid as captured on Oct 9, 2026. Check current pricing on the vendor's site β
Facts
- Platforms
- api (not source-checked)
- API available
- Yes
- Open source
- Yes (not source-checked)
- Model used
- Multi-model (not source-checked)
- Earliest Wayback capture
- 2023-04-25 (earliest archived capture of the site, not a launch date)
Where this tool sits
Use cases
Frequently asked
- How much does Phoenix cost?
- Phoenix operates on a freemium pricing model. Specific pricing details are not documented in the provided information, but it is an open-source platform that can be run locally.
- Can I run Phoenix on my local machine?
- Yes, Phoenix is local-first and can be run on a laptop in under a minute. It is designed to be lightweight for local development and testing environments.
- Does Phoenix work with specific AI models or frameworks?
- Phoenix is vendor-agnostic and works with any model, framework, or language. It supports multi-model usage and integrates with existing tools via native OpenTelemetry support.
- What features does Phoenix offer for AI agents?
- Phoenix traces every step of agent execution, including prompts, retrievals, and tool calls. It allows users to build evaluations to score outputs and run experiments to ship improvements.
- How can I deploy Phoenix at scale?
- Phoenix can be deployed and scaled on Kubernetes using Helm. This allows for robust infrastructure management beyond local laptop usage for production environments.