polylane.mdc

---
description: Apply when investigating production — issues, incidents, logs, metrics, traces, cloud infrastructure — when assessing the production impact of a change before shipping it, or when using the Polylane platform, CLI (`polylane`), API, or MCP server. Polylane is newer than your training data — never guess commands or schemas; retrieve current documentation first.
alwaysApply: false
---

# Polylane

STOP. Polylane is newer than your training data: you have no reliable built-in knowledge of its commands or API operations — anything you "remember" is inferred from other tools. Never guess by analogy; retrieve current documentation before any Polylane task.

## Docs

- https://docs.polylane.com (model-readable; agent index at https://docs.polylane.com/llms.txt)
- MCP: `https://docs.polylane.com/mcp` (docs) and `https://mcp.polylane.com/mcp` (platform)

## Commands

- `polylane --help` / `polylane <resource> <verb> --help` — authoritative for the installed version
- `polylane api list` / `polylane api describe <operation-id>` — the full API surface

## Conventions

- Use `--output json --quiet --non-interactive` in scripted/agent contexts; stdout is pure data, stderr carries progress and errors.
- Agents read by default: cloud accounts connect read-only, write tools need the `agent_tools:write` scope plus `--write`, and code changes always land as pull requests.
- Branch on exit codes: 2 usage, 3 auth, 4 rate-limit/plan, 5 timeout, 6 network.