Best Practices
These patterns come from teams running Long Horizon in production. They'll help you build reliable tests, keep your data secure, and collaborate effectively.
Environment setup
- Standardize your app URL — Document the canonical dev URL for your project. Everyone on the team should use the same port and protocol.
- Match environments — If tests pass locally but fail for teammates, check that feature flags and environment variables match.
- Keep your dev server stable — Don't restart your server mid-test. Long Horizon needs a consistent target.
MCP configuration
- Commit or gitignore deliberately — Decide as a team whether MCP config files should be shared. Document your choice.
- Verify after updates — When you update Long Horizon or your coding agent, confirm MCP is still enabled. Settings can reset.
- Avoid duplicate entries — Don't create multiple MCP server entries for Long Horizon under different names. It causes confusion.
- Restart CLI agents — Claude Code and Codex need a restart after config changes. Cursor needs the toggle re-enabled.
Security
- Default to workspace visibility — Only make reports public when you have a specific external audience.
- Review before publishing — Reports can contain API tokens, user data, and credentials captured in screenshots or network logs.
Team collaboration
- Use the same workspace — Team members should share a workspace so everyone sees the same reports and history.
- Document your rules — Fill in
PLANNING_RULES.mdandTESTING_RULES.mdwith your team's conventions. - Keep skills in sync — When Long Horizon releases updated skills, sync across the team to avoid version mismatches.
- Share reports for code review — Include report links in PRs to show that tests pass.