Open-source standards and tooling for humans and AI agents to draft, validate, and compute legally-meaningful contracts.
Linux Foundation project · Apache-2.0
Accord Project provides the open standard — and reference implementation — for computable contracts: structured, typed, executable agreements that AI agents can reliably read, write, validate, and run.
| Component | What it is | Repo |
|---|---|---|
| Concerto | Schema language for contract data. Generates TypeScript, Java, Go, Python, JSON Schema from one .cto file. |
concerto |
| TemplateMark | Markdown-based contract template format. Variables + conditional logic. Readable by humans and LLMs. | markdown-transform |
| APAP | Agreement Protocol API — REST + MCP server. Agents call contracts as tools. | apap |
Natural language request
│
▼
┌─────────────┐ ┌──────────────┐ ┌──────────────┐
│ TemplateMark│────▶│ Concerto │────▶│ APAP │
│ (template) │ │ (schema) │ │ (MCP / REST) │
└─────────────┘ └──────────────┘ └──────────────┘
│ │ │
Prose text Type-safe data AI tool calls
validation
# Use the template engine API in your project
npm install @accordproject/template-engineTry it live:
- Template Playground — draft and test templates in your browser
- Concerto Playground — paste a schema, see 7 language targets
- Agent Quickstart — end-to-end AI contract drafting demo
Accord Project templates are structured data sources for LLM workflows:
- TemplateMark is Markdown — LLMs already understand it at training-data scale
- Concerto schemas validate LLM-generated data before it reaches the contract
- APAP MCP endpoint exposes templates as tool calls to any MCP client
- TypeScript logic compiles statically — agent-generated code fails at compile time, not runtime
See the AI & Agent Workflows guide and the 2024 Whitepaper.
- Discord — join the Technology Working Group
- Blog — blog & news
- Governance — TSC and working groups
