Builder infrastructure
Arize Phoenix
The open-source platform for agent development and evaluation
Phoenix is Arize's open-source, local-first product for tracing, evaluating, experimenting with, and iterating on AI agents. The captured product page positions it around visibility into agent steps, quality evaluation, datasets and experiments, and a prompt IDE.
Vendor's own claim
What Arize Phoenix is
Vendor's own claimPhoenix is the open-source branch of Arize's AI engineering offering. Arize's site distinguishes it from Arize AX, the managed platform; the retrieved materials say Phoenix is intended for teams that want open-source, self-hosted, or local control.
What it can do
Vendor's own claim- Tracing of agent steps and trajectories
- Evaluations for outputs and failures
- Datasets and experiments for comparing changes
- Prompt iteration through a Prompt IDE
- Self-hosting and local-first use
- PXI, an AI engineering agent inside Phoenix
Who it is for
Vendor's own claimDevelopers and GTM engineering teams building AI agents who want to inspect traces, run evaluations, and control deployment locally or through self-hosting.
When to choose something else
Our readTeams that need a managed SaaS deployment, enterprise support, compliance controls, or longer managed retention should evaluate Arize AX instead: Arize's own product comparison says Phoenix is for open-source/self-hosted/local control while AX is the managed platform. It is also not a no-code sales tool; effective use starts with instrumenting an agent and setting up trace/evaluation work.
Verify before you adopt it
- Phoenix-specific documentation root and installation instructions
- Phoenix API/MCP reference and OpenAPI specification
- Phoenix-specific pricing details
- Status page - vendor name absent from retrieved body
- A machine-readable OpenAPI specification was not included in the reviewed source set.