pytest-receptor

A pytest reporter for coding agents.

When pytest is driven by an agent rather than read by a person, its output is consumed by something that pays for every token and cannot scroll back. pytest-receptor renders the same run for that consumer: it states what happened, groups repeated failures by root cause, and gives the exact command to re-run them.

Note

The 1.x compatibility contract protects the reliability floor, documented outcome meanings, options, and pytest-receptor.events@1. See Compatibility and migration.

Where to go

If you want to…	Read
Install it	Installation
Run it and read its output	Usage
Look up an option, an outcome label, or an output field	Reference
Understand what it does to pytest, and why	How it works
Know which output channel is authoritative	Output channels and authority
Produce or read normalized JSONL evidence	Versioned evidence artifacts
Understand the 1.x promises or migrate from 0.7	Compatibility and migration
Know what it does not do	Limitations
See measured token costs, and how they were measured	Benchmarks

In thirty seconds

$ pytest --receptor=llm

FAIL exit=1 | 38 errors, 90 passed | 12.40s | 1 root cause

[1] TypeError | 38 tests | setup
    conftest.py:31
    TypeError: 'NoneType' object is not subscriptable
    tests:
      tests/test_merge.py::test_merge[0]
      tests/test_merge.py::test_merge[1]
      tests/test_merge.py::test_merge[2]
      +35 more
    rerun: pytest tests/test_merge.py -q

One broken fixture, thirty-eight failing tests, one root cause. Plain pytest spends 3,308 tokens on that run; this is 106.

Installing the plugin changes nothing until you ask it to: the default profile registers nothing at all, so plain pytest behaves exactly as it does without it installed.