For developers and platform teams

Two lines.Every call, token,and dollar.

An import and waxell.init() are the whole integration. Traces land in seconds, priced in real dollars across every provider. Policies are there when you want them, and out of your way when you don't.

Book a demo

30 minutes. No commitment.

SOC 2 Type II attested · ISO/IEC 27001:2022 certified · HIPAA with a BAA · aligned to the EU AI Act and NIST AI RMF

pip install waxell-observe

import waxell_observe as waxell
waxell.init()            # reads WAXELL_API_KEY; every LLM call, tool call and token, traced

@waxell.observe(agent_name="support-bot")   # optional: a named run, with policy checks
def handle_ticket(query: str) -> str: ...

The same five moves, for people who ship

01Discover
Every agent in one registry, including the MCP tools.
Register from the SDK or the CLI. Discover the ones you forgot about from the platforms they run on.
wax agents list. 20 platform adapters.
Go to Discover
02Observe
Every span, token and dollar, per run.
Waxell Observe auto-instruments 160+ Python AI libraries. Open any run and read each model call and tool call in order. Replay a recorded model call against a new prompt version and compare.
OpenTelemetry under the hood. Export anywhere.
Go to Observe
03Enforce
A policy engine you can call from code.
Budgets, PII, egress, tool scope, approvals. Enforced in Runtime or at the MCP Gateway, so the rules don't live in your prompt.
Local gate for tests. Dry-run mode.
Go to Enforce
04Run
Push with one command, then tail and schedule it.
Two decorators put a workflow under Runtime: a policy check before every step, durable checkpoints, and a budget that stops the run and records why.
wax push · wax runs tail · 10 CPU-hour Runtime trial (30 days)
Go to Run
05Prove
The audit trail is a side effect of running.
Every decision is a record. Compliance stops being a quarterly scramble for the platform team.
Evidence packs from the same spans you already emit.
Go to Prove

Start here

Plays for this team

All plays

Two lines.Every call, token,and dollar.

Every agent in one registry, including the MCP tools.

Every span, token and dollar, per run.

A policy engine you can call from code.

Push with one command, then tail and schedule it.

The audit trail is a side effect of running.