Why WizardFlow?
One agent run shouldn’t require an observability platform.
WizardFlow turns an agent run into a portable JSONL file you can replay locally or drop into the browser.
When one run needs a closer look
You want to show a teammate why one run went wrong. Traditional observability products are designed around persistent infrastructure: servers, accounts, instrumentation pipelines, and dashboards.
Those capabilities are valuable in production—but often excessive when the task is simply: “Look at this run.”
The trace is the artifact
A WizardFlow trace is an ordinary file. You can:
- attach it to a bug report;
- send it to a teammate;
- commit or archive it;
- inspect or diff it with ordinary tools;
- preserve representative runs as benchmarking artifacts;
- replay it without connecting to a trace backend.
Two ways to replay
On your machine
wizardflow ui run.jsonlopens the complete viewer locally.
In the browser
Non-developers can drop the same file into getwizardflow.com. The hosted viewer parses it in the browser; the trace is not uploaded.
What you avoid
- No account
- No trace server or database
- No runtime dependencies
- No framework requirement
- No permanent platform commitment
A focused tool, not a smaller platform
| Capability | WizardFlow | Observability platforms |
|---|---|---|
| Primary artifact | Portable JSONL file | Stored platform trace |
| Local replay | Included | Usually platform-oriented |
| Production monitoring | Not the focus | Core capability |
| Team dashboards and evaluations | Not the focus | Core capability |
Where platforms win
If you need centralized production monitoring, shared dashboards, hosted evaluations, alerting, or long-term team analytics, a full observability platform may be the better tool.
WizardFlow is designed for recording, replaying, and sharing individual runs as files.