vBRIEF (Basic Relational Intent Exchange Format) is a universal format for structured thinking — from a quick todo list to a full project plan to an AI agent's memory. One schema, graduated complexity, token-efficient by design.
TLDR: Every AI agent invents its own memory format. Every planning tool has its own schema. vBRIEF is the common language — an open format that unifies todos, plans, playbooks, specs, and agent memory into a single Plan model. Start with 4 fields, graduate to DAG workflows. Token-efficient by design.
A minimal vBRIEF document has just four fields:
{
"vBRIEFInfo": { "version": "0.5" },
"plan": {
"title": "My First Plan",
"status": "running",
"items": [
{ "title": "Do the thing", "status": "pending" }
]
}
}That's a valid vBRIEF document. Everything else is optional.
Start simple. Add structure only when you need it.
- Minimal — A flat task list. Title, status, items. →
examples/minimal-plan.vbrief.json - Structured — Add narratives for context and rationale. →
examples/structured-plan.vbrief.json - Retrospective — Capture outcomes, strengths, weaknesses, lessons. →
examples/retrospective-plan.vbrief.json - Graph / DAG — Add edges for dependencies and workflows. →
examples/dag-plan.vbrief.json
- Token efficient — TRON encoding cuts LLM token usage by 35–40%
- DAG support — Model dependencies, pipelines, and conditional workflows
- Graduated complexity — No boilerplate; add features only as needed
- Interoperable — JSON Schema validation, standard JSON/TRON serialization
- Open standard — RFC-style specification, no proprietary extensions
- No vendor lock-in — Plain files, any tool can read/write them
| Document | Description |
|---|---|
| vbrief-spec-0.5.md | Formal specification (RFC 2119) |
| GUIDE.md | Reference guide with patterns and recipes |
| docs/getting-started.md | Tutorial for beginners |
| docs/tron-encoding.md | TRON format reference |
| MIGRATION.md | v0.4 → v0.5 migration guide |
vBRIEF/
├── vbrief-spec-0.5.md # Formal specification
├── GUIDE.md # Reference / cookbook
├── examples/ # Graduated complexity examples (JSON + TRON)
├── schemas/ # JSON Schema
├── libvbrief/ # Python library
├── docs/ # Guides and references
├── validation/ # Validators
├── tests/ # Test suite
└── history/ # Archived drafts and old docs
pip install libvbriefOr from source:
git clone https://github.com/visionik/vBRIEF.git
cd vBRIEF
pip install -e .python validation/vbrief_validator.py your-plan.vbrief.jsonSee CONTRIBUTING.md. Feedback and issues welcome at GitHub Issues.
Open standard. See repository for license details.