AI-maintained MEV block builder, forked from flashbots/rbuilder.
| Phase | Title | Status | Plan |
|---|---|---|---|
| 1 | Foundation | completed | 000 - Origin Plan |
| 2 | CI/CD Upgrade | completed | 000 - Origin Plan |
| 3 | Bot Deployment | completed | 000 - Origin Plan |
| 4 | Testing Harness | completed | 000 - Origin Plan |
| 5 | Self-Improvement | completed | 000 - Origin Plan |
- Plans Index
- Session Logs
- Agent Instructions: CLAUDE.md
- Upstream: flashbots/rbuilder
- 001 - Integration Test Scenarios — bundle inclusion, cancellation, bid value tests
- Upstream sync — first sync with flashbots/rbuilder (see
.ai/upstream-sync.md) - Agent-driven issues — label issues with
claudeoragent-safeto start autonomous development
When making implementation decisions, prefer:
- Safety first - Never compromise relay submission guards, blocklist checks, or key management
- Upstream compatibility - Minimize divergence from flashbots/rbuilder
- AI-friendly - Optimize for agent comprehension (explicit over implicit, tables over prose)
- Incremental - Each PR should be independently reviewable and revertable
vbuilder/
├── PLAN.md # This file - project dashboard
├── CLAUDE.md # Agent instructions
├── docs/
│ ├── plans/ # Implementation plans
│ └── sessions/ # Agent session logs
├── .ai/ # Agent context docs
│ ├── architecture.md # LiveBuilder pipeline and crate map
│ ├── testing.md # 3-tier testing strategy
│ ├── safety.md # Safety-critical paths and rules
│ └── upstream-sync.md # Upstream sync strategy and divergence tracking
├── .claude/commands/ # Custom slash commands (test, lint, new-plan, session-log)
├── crates/ # Rust workspace crates
├── scripts/ # Build and test scripts
└── .github/workflows/ # CI/CD workflows