Trellis

Trellis is an open-source AI coding harness built for teams that want reliable, repeatable AI-assisted development at scale. It replaces bloated CLAUDE.md, AGENTS.md, and .cursorrules files with a structured .trellis/ directory of scoped specs, task PRDs, workflow gates, and session memory. Trellis works across multiple AI coding platforms — including Claude Code, Cursor, Codex, OpenCode, and Pi Agent — giving every agent the same project source of truth without stuffing the entire codebase into every prompt.

• Progressive Spec System — Organizes team standards and coding guidelines into scoped spec files under .trellis/spec/ that agents load only when relevant, keeping prompts lean.
• Task PRDs & Context — Stores task requirements, acceptance criteria, review notes, and status in .trellis/tasks/ so agents always have the right task boundary.
• Shared Workflow Lifecycle — Defines a repeatable plan → build → check → finish → learn loop via .trellis/workflow.md, enforcing consistent quality gates across the team.
• Session Memory & Journals — Persists developer decisions, handoff notes, and session context in .trellis/workspace/ so the next agent starts with full context.
• Multi-Platform Adapters — Generates platform-specific entry points (commands, hooks, skills, prompts) for each supported AI coding tool automatically.
• Skill-First Architecture — Routes work through skills for brainstorming, spec loading, implementation checks, and knowledge capture; on hook-capable platforms, context loads automatically.
• Easy Installation — Install globally via npm install -g @mindfoldhq/trellis@beta, then run trellis init -u your-name inside any repository to get started.
• Team & Solo Use — Solo developers gain memory and repeatable workflow; teams gain shared standards, task boundaries, reviewable context, and platform portability.
• AGPL-3.0 Licensed — Fully open-source under the GNU Affero General Public License v3.0, free to use, modify, and distribute.