Open standard · Apache-2.0 · agent-independent
Let AI-built software
grow without sprawling.
Generating code is easy now. Keeping it under control is not. Trellis is a capability-first, evidence-governed standard that keeps AI-generated software understandable, testable and maintainable as it grows - on top of the agent you already use.
- Claude Code
- Codex
- Copilot
- Cursor
- Windsurf
- Gemini
The problem Trellis solves
Spec tools tell an agent what to build. Trellis preserves why it was built that way, how each piece is verified, how much it can grow, and when it must be reconsidered.
Unbounded growth
Duplicated responsibilities, runaway dependencies, ballooning files.
Architecture drift
Locally sensible changes that are globally inconsistent.
Lost rationale
Decisions live in chat history, not in the repository.
Hidden tests
Only engineers can see whether a feature actually works.
Nine pillars, one standard
One-prompt setup
Paste this into your agent. It fetches the Trellis spec, reads your project, scaffolds governance, and discards the install file when done. No downloads, no wizard required.
Fetch https://raw.githubusercontent.com/e-onux/trellis/main/INSTALL.md and follow its instructions.
Need to pick a profile, preset or specific pillars? Use the wizard →
Wizard
Pick your setup. Get a tailored AI spec to paste into your agent, a reproducible npx command, or a ready-to-use .zip - all generated by the same engine the CLI uses.