Dignite Vault Extract MCP server
Commit Nov 15, 2026Converts scans, photos, and office files into structured Markdown and metadata for RAG platforms and AI clients via MCP.
Also known as dignite-projects/vault
- Stars
- 7
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 15, 2026
- License
- LGPL-3.0
What is Dignite Vault Extract MCP?
Dignite Vault Extract converts scans, photos, and office files into structured Markdown and metadata for RAG platforms and AI clients via MCP.
Summary generated from the project's README
About Dignite Vault Extract MCP server
Vault acts as a specialized channel layer designed to transform unstructured content into trustworthy data. It processes scans, photos, PDF images, and Office files to produce structured Markdown and metadata. This service is built for developers and businesses that need to feed high-quality information into RAG platforms, business systems like ERP or HR, and AI clients such as Claude Desktop or Cursor.
To operate, the server requires an OpenAI-compatible LLM provider, which can be a cloud service or a local instance via Ollama. Users must configure an API key and specify vision-capable models for the default OCR provider. The system is distributed as an ABP application module, allowing for integration into existing .NET environments via NuGet packages or through a standalone host application. It utilizes tools like PaddleOCR via a Docker sidecar to assist with text extraction tasks.
What you can do with it
- 1Automate field extraction from documentsExtract specific type-bound fields from digital or scanned documents using vision-capable LLMs.
- 2Supply data to AI desktop clientsProvide structured document data directly to MCP-compatible clients like Cursor or Claude Desktop.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachabledignite-projects/vault is public and reachablegithub.com/dignite-projects/vault ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)LGPL-3.0spdx.org/licenses/LGPL-3.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-13github.com/dignite-projects/vault ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubDignite Vault Extract
Dignite Vault Extract = any content requiring IDP (Intelligent Document Processing) — scans / photos / PDF images / Office files / digital-born documents → trustworthy structured data. A channel layer, not an end-product. It doesn't consume, doesn't own, doesn't dive into business — it hands Markdown + structured metadata to downstream RAG platforms, business systems, and AI clients via REST / EventBus / MCP server / Webhook (planned).
For the positioning, architecture, what the module does not contain, the event contract and the security rules for LLM call paths, see CLAUDE.md. It is the truth source — this README only stages the entry points; everything operational lives in docs/.
Data flow
content requiring IDP: scans / photos / PDF images / Office files / digital-born documents
↓
[Dignite Vault Extract channel]: OCR + Markdown + system metadata + type-bound field extraction
↓ (REST / EventBus / MCP server / Webhook — planned)
├─→ downstream RAG platform
├─→ business systems (finance / CLM / HR / ERP)
├─→ AI clients (Claude Desktop / Cursor / any MCP client)
└─→ any consumer (build your own subscriber)
Solution structure
vault/
├── modules/
│ ├── vault.parse/ # Vault Parse — the text-extraction library: OCR + Markdown providers behind ITextExtractor
│ └── vault.extract/ # Channel implementation — ABP layers (Abstractions / Domain.Shared / Domain / Application / EntityFrameworkCore / HttpApi / Mcp)
│ # and angular/ (the @dignite/ng.vault-extract library)
├── src/ # Host application — provider wiring (OCR + AI) and middleware (ASP.NET Core API)
├── angular/ # Host SPA (operator UI); compiles the module libraries from source
├── etc/ # Docker, ABP Studio run profile, scripts, PaddleOCR sidecar
└── docs/ # Documentation: setup, configuration, egress contracts, deployment
Business modules (contract management / invoice management / HR records / etc.) are not in this repo — they belong on the downstream consumer side per the channel philosophy.
Use it as a module
Vault Extract is distributed as an ABP application module: the Dignite.Vault.Extract.* NuGet packages for the server and the @dignite/ng.vault-extract npm package for the Angular operator UI. This repository's host (src/Dignite.Vault.Extract.Host/) is one host built from them. To add the modules to an ABP application of your own, see Installation.
Quick start (run this repository)
| Requirement | Version |
|---|---|
| .NET SDK | 10.0 |
| Node.js | ^22.22.3 or ^24.15.0 (Angular 22) |
| SQL Server | 2019+ (LocalDB works for development) |
| ABP CLI | matching ABP 10.7 |
| An LLM provider | mandatory — any OpenAI-compatible endpoint; a local Ollama works |
| Docker Desktop | optional — PaddleOCR sidecar and the local OpenTelemetry dashboard |
1. Configure the LLM provider. Create src/Dignite.Vault.Extract.Host/appsettings.secrets.json. It is git-ignored and never published; do not put credentials in the committed appsettings.json or appsettings.Development.json.
{
"Vault": {
"Extract": {
"Endpoint": "https://api.openai.com/v1",
"ApiKey": "sk-...",
"ChatModelId": "gpt-6-luna",
"VisionOcrModelId": "gpt-6-luna",
"ReasoningEffort": "None"
}
}
}
The host fails fast at startup while ApiKey is still the committed placeholder, because classification and field extraction have no non-LLM fallback. VisionOcrModelId must name a vision-capable model: the host's default OCR provider is Vision-LLM. See AI Provider.
2. Run the backend.
cd src/Dignite.Vault.Extract.Host
abp install-libs
dotnet run --migrate-database # first run, and after pulling a new migration; seeds admin / 1q2w3E*
dotnet run
API: https://localhost:44348. Swagger: https://localhost:44348/swagger.
3. Run the Angular UI.
npm install # once, at the repo root: one install for every Angular workspace
cd angular
npm start
SPA: http://localhost:4200. Sign in with admin / 1q2w3E*.
The full walkthrough — database connection, sidecars, proxy regeneration, tests, troubleshooting — is in Local Development.
Where to go next
Start at the documentation index.
| You want to… | Read |
|---|---|
| Run this repository on your machine | Local Development |
| Add the modules to your own ABP application | Installation |
| Define what gets extracted | Document Types and Fields |
| Choose an OCR provider | Text Extraction |
| Integrate a downstream system | Integration Events, REST API |
| Connect an AI client | MCP Server |
| Deploy to production | Deployment, Deployment Checklist |
Release notes, including breaking changes and their migrations, are in CHANGELOG.md.
External references:
License
Dignite Vault Extract is licensed under the GNU Lesser General Public License v3.0.
Dignite Vault Extract MCP questions
What kind of files can the server process?
The server processes scans, photos, PDF images, Office files, and digital-born documents.
How can I integrate this into my own ABP application?
You can add the modules to your own ABP application using the Dignite.Vault.Extract NuGet packages for the server and the @dignite/ng.vault-extract npm package for the Angular UI.