Turn meeting transcripts into executable smart contracts in seconds.
SESAP is an AI-powered platform that transforms unstructured meeting transcripts into structured, executable Smart Social Contracts using the Accord Project technology stack. It features a browser-based Template Playground for designing, testing, and previewing contracts in real-time — and an MCP server that brings contract generation directly into GitHub Copilot Chat.
At its core, SESAP is an experiment in building a foundation for autonomous economic execution. Turning natural language into rigorous Accord Project schemas is the hardest part. Once a conversation is structured into a strictly typed .cto model, it unlocks the ability for an AI Workforce to operate on your behalf.
Imagine a network of reasoning agents that:
- Evaluate the terms of the Smart Social Contract.
- Execute workflows based on those terms to generate revenue for the user.
- Sign and govern agreements autonomously, held accountable by underlying economic incentives and structured evaluations.
SESAP provides the "DocuSign for Agents" layer, bridging the gap between human intent (meeting transcripts) and autonomous digital economies.
- Initial Poly186 repository setup and open-source launch
- UI execution frame for agent evaluation and signing concepts
- Connect SESAP with Poly186 Identity and Authentication layers
- Enable one-click contract deployment to Poly186 ledgers
- Seamless integration with the broader Poly186 Agent Network
┌──────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌──────────────┐
│ Meeting │ │ Step 1: GPT-5.1 │ │ Step 2: GPT-5-mini│ │ Step 3: │
│ Transcript │────▶│ Extract structure │────▶│ Generate Accord │────▶│ Validate & │
│ (plain text) │ │ (parties, scope, │ │ artifacts (model, │ │ polish │
│ │ │ timeline, terms) │ │ template, data) │ │ artifacts │
└──────────────────┘ └────────────────────┘ └────────────────────┘ └──────┬───────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ Accord Engine: Concerto Model + TemplateMark → HTML Preview │
└──────────────────────────────────────────────────────────────┘
- Paste a transcript — Upload or paste a meeting transcript
- AI extracts & generates — A 3-step GPT pipeline extracts contract structure, generates Accord Project artifacts (Concerto model, TemplateMark template, JSON data), and validates them
- Edit & preview — Use Monaco editors to refine the model, template, and data. See the rendered HTML agreement update in real-time
- AI Contract Generation — 3-step LLM pipeline (GPT-5.1 → GPT-5-mini → GPT-5-mini) with Zod-validated structured output
- Template Playground — Browser-based editors for Concerto models (.cto), TemplateMark templates (.tem.md), and JSON data
- Real-time Preview — Instant HTML agreement rendering via the Accord Project engine
- MCP Server — Use SESAP tools directly in GitHub Copilot Chat (generate, render, validate contracts)
- Share Links — Compress and share contract configurations via URL
- Dark/Light Theme — Full theme support with persistence
┌──────────────────────────────────────────────────────────────────┐
│ SESAP Architecture │
├──────────────────────────────────────────────────────────────────┤
│ │
│ Frontend (Vite 4 / React 18) Backend (Express 5) │
│ ├── Zustand (state management) ├── POST /api/generate │
│ ├── Monaco Editor (CTO, MD, JSON) ├── Azure AI Foundry │
│ ├── Ant Design 5 (UI components) │ ├── GPT-5.1 (heavy) │
│ ├── Accord Engine (browser-side) │ ├── GPT-5-mini (med) │
│ └── React Router (SPA) │ └── GPT-5-nano (light) │
│ ├── Accord Engine (server) │
│ proxy /api/* ──────────────▶ └── Zod 4 (schemas) │
│ │
│ MCP Server (stdio) │
│ ├── generate_contract (transcript → artifacts + HTML) │
│ ├── render_contract (model + template + data → HTML) │
│ └── validate_contract (model + template + data → errors) │
│ │
└──────────────────────────────────────────────────────────────────┘
- Node.js ≥ 18
- npm ≥ 6
- Azure AI Foundry account (for AI contract generation)
git clone https://github.com/Poly186-AI-DAO/sesap.git
cd sesap
npm installcp .env.example .envEdit .env with your Azure AI Foundry credentials:
AZURE_AI_FOUNDRY_BASE_URL=https://your-resource.cognitiveservices.azure.com
AZURE_AI_FOUNDRY_KEY=your-api-key
AZURE_GPT_5_1_DEPLOYMENT=gpt-5.1
AZURE_GPT_5_MINI_DEPLOYMENT=gpt-5-mini
AZURE_GPT_5_NANO_DEPLOYMENT=gpt-5-nanoTemplate Playground (no API key needed for manual editing):
npm run devWith AI Contract Generation (requires Azure credentials):
# Terminal 1: Start the API server
npm run server
# Terminal 2: Start the frontend
npm run devnpm run buildSESAP includes an MCP (Model Context Protocol) server that integrates directly with GitHub Copilot Chat in VS Code.
Add to your .vscode/mcp.json (already configured in this repo):
{
"servers": {
"sesap": {
"type": "stdio",
"command": "npx",
"args": ["tsx", "server/mcp/index.ts"],
"env": {
"AZURE_AI_FOUNDRY_BASE_URL": "${env:AZURE_AI_FOUNDRY_BASE_URL}",
"AZURE_AI_FOUNDRY_KEY": "${env:AZURE_AI_FOUNDRY_KEY}",
"AZURE_GPT_5_1_DEPLOYMENT": "${env:AZURE_GPT_5_1_DEPLOYMENT}",
"AZURE_GPT_5_MINI_DEPLOYMENT": "${env:AZURE_GPT_5_MINI_DEPLOYMENT}",
"AZURE_GPT_5_NANO_DEPLOYMENT": "${env:AZURE_GPT_5_NANO_DEPLOYMENT}"
}
}
}
}| Tool | Description |
|---|---|
generate_contract |
Transcript → full pipeline → Accord artifacts + HTML |
render_contract |
Model + template + data → rendered HTML |
validate_contract |
Model + template + data → validation result with errors |
| Resource | URI | Description |
|---|---|---|
| Sample Template | sesap://samples |
Playground sample with Concerto model and TemplateMark |
| Platform Info | sesap://info |
SESAP capabilities and pipeline description |
@workspace Use the SESAP render_contract tool to render this template
with the sample data from sesap://samples
| Layer | Technology | Version |
|---|---|---|
| Frontend | React + TypeScript | 18.x / 5.x |
| Build | Vite | 4.x |
| State | Zustand (immer + devtools) | 4.x |
| UI | Ant Design + styled-components | 5.x |
| Editors | Monaco Editor | 0.50.x (CDN) |
| Contract Engine | Accord Project (Concerto, TemplateMark, CiceroMark) | 3.20.x / 2.3.x |
| LLM | Azure AI Foundry via OpenAI SDK | 6.x |
| Schema Validation | Zod | 4.x |
| Backend | Express | 5.x |
| MCP | @modelcontextprotocol/sdk | 1.x |
| Testing | Vitest + React Testing Library | 1.x |
sesap/
├── src/ # Frontend source
│ ├── App.tsx # App initialization and routing
│ ├── main.tsx # Entry point, Monaco config
│ ├── store/store.ts # Zustand state management + Accord rebuild
│ ├── editors/ # Monaco editor components (CTO, MD, JSON)
│ ├── components/ # UI components (Navbar, TranscriptUpload, etc.)
│ ├── samples/ # Playground sample data
│ └── tests/ # Unit tests (Vitest)
├── server/ # Backend source
│ ├── api.ts # Express API server (port 3001)
│ ├── accord/engine.ts # Server-side Accord rendering
│ ├── llm/azure-client.ts # Azure OpenAI client with retry logic
│ ├── schemas/contract.ts # Zod schemas for structured LLM output
│ ├── scripts/ # 3-step LLM pipeline
│ └── mcp/ # MCP server (stdio transport)
│ ├── index.ts # Entry point
│ ├── tools.ts # Contract tools (generate, render, validate)
│ └── resources.ts # Resources (samples, platform info)
├── docs/ # Documentation
├── .env.example # Environment variable template
└── .vscode/mcp.json # MCP server configuration
npm testRuns 53 unit tests covering:
- Zod schema validation (all contract schemas)
stripNullValuesutility (Concerto compatibility)- Store utility functions (error formatting, theme, debounce cancelation)
- UI components (Navbar rendering)
- Compression utilities (share link encoding/decoding)
| Script | Command | Description |
|---|---|---|
dev |
npm run dev |
Start Vite dev server |
server |
npm run server |
Start Express API server |
mcp |
npm run mcp |
Start MCP server (stdio) |
build |
npm run build |
Production build |
test |
npm test |
Run unit tests |
See CONTRIBUTING.md for guidelines on contributing to SESAP.
Apache License 2.0 — Copyright © 2024-2026 Poly186
Built with the Accord Project and Azure AI Foundry