π¦
Agentic Diligence Ibanking
AI-first agentic M&A due diligence platform powered by GitHub Copilot SDK
0 installs
Trust: 34 β Low
Agents
Ask AI about Agentic Diligence Ibanking
Powered by Claude Β· Grounded in docs
I know everything about Agentic Diligence Ibanking. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Loading tools...
Reviews
Documentation
Agentic Diligence β AI-First M&A Due Diligence
Overview
An AI-first due diligence platform that autonomously performs first-pass M&A analysis. Powered by the GitHub Copilot SDK, it orchestrates specialist tools across public research, private data rooms, financial analysis, and legal review β producing IC-quality deliverables in minutes.
Key Features
- Autonomous Diligence Agent β Single orchestrator drives 5-phase analysis workflow using 17 MCP tools
- Multi-Deal Pipeline β Run diligence on multiple targets simultaneously (Atlas, Titan, Meridian)
- IC-Quality Deliverables β Generates investment memo (PDF), summary deck (PPTX), and interactive dashboard
- Real-Time Observability β Watch every tool call, finding, and decision in a live event stream
- Evidence-Based Analysis β Every claim traces to a source with provenance labels and confidence scores
- Contradiction Detection β Automatically flags discrepancies between management claims and actual data
- Human-in-the-Loop β Approval gates for sensitive actions (seller questions), steering for mid-run guidance
- Session Persistence β SQLite locally, pluggable for Azure SQL in production
- Artifact Storage β Local filesystem or Azure Blob Storage
Architecture
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β Next.js Frontend (:3001) β
β Pipeline β Timeline β Intel Panel β Dashboard Modal β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β Express API Server (:3000) β
β /run β /events (SSE) β /artifacts β /steer β /deals β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β GitHub Copilot SDK (Agent Loop) β
β Orchestrator β Hooks β Skills β Event Bridge β
ββββββββ¬βββββββ¬βββββββ¬βββββββ¬βββββββ¬βββββββββββββββββββββββ€
β Web β VDR β Fin β Work β Memo β Artifacts β
β Rsrchβ β ance β flow β β β
β MCP β MCP β MCP β MCP β MCP β MCP β
ββββββββ΄βββββββ΄βββββββ΄βββββββ΄βββββββ΄βββββββββββββββββββββββ
Tech Stack
| Component | Technology |
|---|---|
| Agent Runtime | GitHub Copilot SDK (@github/copilot-sdk) |
| LLM | Azure AI Foundry (GPT-5.4-mini) |
| MCP Servers | @modelcontextprotocol/sdk (6 stdio servers, 17 tools) |
| Frontend | Next.js 16, React 19, Tailwind CSS |
| API Server | Express with SSE streaming |
| Artifact Gen | PDFKit (memo), PptxGenJS (deck), Next.js (dashboard) |
| Session Store | SQLite (local) / Azure SQL (production) |
| Artifact Store | Filesystem (local) / Azure Blob Storage (production) |
Quick Start
Prerequisites
- Node.js 18+ and pnpm 9+
- GitHub Copilot CLI installed and authenticated
- Azure AI Foundry endpoint (or compatible OpenAI API)
Setup
# Clone the repo
git clone https://github.com/jawadaminmsft/agentic-diligence-ibanking.git
cd agentic-diligence-ibanking
# Install dependencies
pnpm install
# Configure environment
cp .env.example .env
# Edit .env with your Azure AI Foundry endpoint and auth
# Build all packages
pnpm --filter mcp-servers build
pnpm --filter harness build
# Start the API server
npx tsx packages/harness/src/serve.ts
# In another terminal, start the frontend
cd packages/web && npx next dev --port 3001
Run Diligence
- Open http://localhost:3001
- Select a deal from the pipeline (Atlas, Titan, or Meridian)
- Click "Run Diligence" β the agent starts autonomous analysis
- Watch real-time events in the timeline
- Review memo, issues, and evidence in the Intelligence panel
- View generated artifacts (PDF memo, PPTX deck, interactive dashboard)
Copilot SDK Features Used
- Agent Loop β Multi-turn LLM β tool call orchestration
- MCP Servers β 6 domain-specific tool servers with 17 tools
- Skills β 9 SKILL.md prompt modules for domain expertise injection
- Hooks β Pre/post tool-use hooks for approval gates and provenance tagging
- Streaming Events β Real-time event bridge from SDK to frontend via SSE
- Session Persistence β Structured session IDs with SQLite/Azure SQL storage
- BYOK Provider β Azure AI Foundry with Azure AD token authentication
Project Structure
packages/
βββ harness/ # Core runtime β Copilot SDK integration
β βββ src/agents/ # Orchestrator prompt
β βββ src/hooks/ # Approval gate, provenance tagger, lifecycle
β βββ src/skills/ # 9 SKILL.md domain expertise modules
β βββ src/events/ # Event bridge, types, trace store
β βββ src/db/ # Session persistence (SQLite/Azure SQL)
β βββ src/storage/ # Artifact storage (local/Azure Blob)
β βββ src/server.ts # Express API server
βββ mcp-servers/ # 6 MCP tool servers
β βββ src/web-research/ # Public source search
β βββ src/vdr/ # Virtual data room
β βββ src/finance/ # Financial analysis
β βββ src/workflow/ # Issue tracking
β βββ src/memo/ # Memo section management
β βββ src/artifacts/ # PDF, PPTX, dashboard generation
βββ web/ # Next.js operator UI
β βββ app/ # Pages (landing, dashboard)
β βββ components/ # UI components
β βββ hooks/ # Event stream, run state
βββ data/ # Synthetic deal corpus (not committed)
Documentation
- Business Use Cases β Target users, workflow, value proposition
- Technical Architecture β System design, SDK integration, data flows
License
This project is licensed under the MIT License. This is a demonstration and reference implementation β not an official Microsoft product. See LICENSE for full terms.
