MCP server de AskAi
Commit 6 nov 2026Provides a modular RAG platform for document indexing, hybrid search, summarization, and conversational AI via chat, REST, or MCP.
También conocido como FaastLab/AskAi
- Stars
- 1
- Herramientas
- —
- Descargas, 30 días
- —
- Última versión
- —
- Último commit
- 6 nov 2026
- Licencia
- Apache-2.0
¿Qué es AskAi MCP?
Provides a modular RAG platform for document indexing, hybrid search, summarization, and conversational AI via chat, REST, or MCP.
Resumen generado a partir del README del proyecto
Acerca del MCP server de AskAi
Building a knowledge layer for AI agents or providing document intelligence for human teams is what AskAi enables. This platform manages the entire document lifecycle by ingesting files, parsing them, and creating embeddings for retrieval. It supports hybrid search using vector, keyword, and metadata components alongside reranking to ensure high-quality results. Users can interact with their data through a chat interface, a REST API, a Python SDK, or this MCP server.
Developers can deploy the system locally using Docker with a default stack of MinIO, Postgres with pgvector, and OpenAI. For enterprise environments, the architecture uses an adapter layer that allows you to swap components for Azure or AWS services like Azure AI Search or S3 without changing your application code. This flexibility makes it suitable for self-hosting or running in regulated environments where data must stay on-premises or in specific cloud regions.
Qué puedes hacer con él
- 1Provide knowledge to AI agentsConnect an agentic AI framework to a centralized document repository using the MCP server or REST API.
- 2Automate regulatory compliance validationCreate validator agents that check internal company reports against a corpus of regulatory handbooks and policy papers.
- 3Summarize long documentsUse map-reduce techniques to generate summaries for documents regardless of their length.
- 4Chat with internal company documentationEnable employees to ask questions about documents stored in SharePoint, Confluence, or S3 through a chat UI.
Escrito a partir del README del proyecto. Cada caso de uso se verifica con una cita del mismo.
Verificaciones automáticas
Ejecutado 16 nov 2026Informe completo →Aún no verificado. Un servidor necesita una puntuación de 80, ningún hallazgo grave y suficientes datos para que podamos revisarlo. Cómo funcionan las verificaciones · Qué corregir para obtener más puntos
- Repositorio de origen vinculado y accesiblefaastlab/askai is public and reachablegithub.com/faastlab/askai ↗Aprobado
- El paquete indica este repositorio como su origenNo verificado
- Procedencia de la compilación (atestación de Sigstore)No verificado
- Licencia de código abierto (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Aprobado
- Commit o lanzamiento en el último añoLast commit or release 2026-11-06github.com/faastlab/askai ↗Aprobado
- OpenSSF ScorecardNo verificado
- El nombre del paquete no es similar a un paquete popularNo verificado
- No hay avisos críticos en el paquete ni en sus dependenciasNo verificado
- Sin instrucciones ocultas para el modelo0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersAprobado
- Las herramientas destructivas están etiquetadasNo verificado
- El endpoint remoto responde al handshake de MCPNo verificado
README
Ver en GitHubFaastLab AskAi
The open knowledge platform for humans and AI agents. Ingest documents once. Query via chat, REST, MCP, or SDK. Deploy fully open-source, on Azure, on AWS, on-prem, or hybrid.
What is AskAi?
AskAi is FaastLab's flagship open-source platform for document intelligence. It does four things, all production-grade, all modular:
- Indexing — ingest documents from any source, parse, chunk, embed
- Search & RAG — hybrid retrieval (vector + keyword + metadata) with reranking
- Summarisation — map-reduce summaries for documents of any length
- Ask AI — chat over your documents, with citations and multi-step reasoning
It is designed to serve three audiences from one codebase:
- Humans — via a chat UI (Ask AI)
- AI agents — via REST, MCP server, or Python SDK (knowledge layer for agentic AI)
- Validators — agents that read documents and validate them against a corpus (e.g. regulatory report compliance)
Why AskAi exists
Most open-source RAG projects are demos. Most enterprise knowledge platforms are SaaS-only or vendor-locked. There's a gap in the market for an open, modular, agent-ready knowledge platform that companies can:
- Self-host fully open-source
- Deploy on Azure / AWS / GCP using managed services
- Mix and match (e.g. OpenAI for LLM, Postgres for vectors, S3 for storage)
AskAi fills that gap. Every external dependency is behind an adapter — swap OpenAI for Azure OpenAI, swap Postgres for Azure AI Search, swap MinIO for S3 — without changing application code.
Architecture at a glance
┌────────────────────────────────────────────────────────────────┐
│ Interfaces (pick any/all) │
│ Chat UI (Next.js) │ REST API │ MCP Server │ Python SDK │
└────────────────────────────────────────────────────────────────┘
│
┌────────────────────────────────────────────────────────────────┐
│ Core Modules │
│ ┌──────────┐ ┌────────────┐ ┌──────────────┐ ┌──────────┐ │
│ │ Indexing │→ │ Search/RAG │→ │ Summarisation│→ │ Ask AI │ │
│ └──────────┘ └────────────┘ └──────────────┘ └──────────┘ │
└────────────────────────────────────────────────────────────────┘
│
┌────────────────────────────────────────────────────────────────┐
│ Adapter Layer (swappable) │
│ Storage │ Vector DB │ LLM │ Embeddings │ Queue │ Auth │
└────────────────────────────────────────────────────────────────┘
│
┌────────────────────────────────────────────────────────────────┐
│ Default OSS stack (works out of the box) │
│ MinIO │ Postgres+pgvector │ OpenAI │ Redis+Celery │ JWT │
│ │
│ Azure stack (one config change) │
│ Blob │ Azure AI Search │ Azure OpenAI │ Service Bus │ Entra │
└────────────────────────────────────────────────────────────────┘
See Architecture.md for the full design.
Quick start
# Clone and enter
git clone https://github.com/faastlab-ai/askai.git
cd askai
# Set OpenAI key (or Azure OpenAI — see .env.example)
cp .env.example .env
# Spin up everything
docker compose up -d
# Ingest a sample corpus (UK Financial Regulation)
make demo-corpus
# Open the chat UI
open http://localhost:3000
That's it. You now have a working knowledge platform with chat, REST API, and MCP server — all running locally, no cloud account needed.
Use cases
For companies wanting chat over docs Drop in your SharePoint, Confluence, or S3 documents. Get a chat UI for your team.
For companies building AI agents Use AskAi as the knowledge backbone. Any agent (Claude, LangGraph, CrewAI, custom) can call it via REST or MCP.
For fintech and regulated firms Ingest the regulatory handbooks, supervisory statements, and policy papers your firm has to comply with (FCA, PRA, BoE, equivalents). Build validator agents that check internal reports against the corpus, with paragraph-level citations back to the source rule.
Deployment options
| Mode | Storage | Vector DB | LLM | Use case |
|---|---|---|---|---|
| Pure OSS | MinIO | Postgres+pgvector | OpenAI / local LLM | Self-host, no cloud lock-in |
| Azure | Blob | Azure AI Search | Azure OpenAI | Enterprise on Azure |
| AWS | S3 | OpenSearch / pgvector on RDS | Bedrock / OpenAI | Enterprise on AWS |
| Hybrid | S3 + Postgres on-prem | pgvector | Azure OpenAI | Mixed regulated environments |
All controlled by config. No code changes between modes.
Project status
Active development. See Architecture.md for the full design.
License
Apache License 2.0 — free for any use, commercial or otherwise, with an explicit patent grant. Built and maintained by FaastLab.Ai.
Preguntas sobre AskAi MCP
What can AskAi do?
It performs document indexing, hybrid search and RAG, summarization, and provides a chat interface for querying documents with citations.
How can AI agents interact with AskAi?
AI agents can use AskAi as a knowledge layer by connecting through a REST API, an MCP server, or a Python SDK.
Can I deploy AskAi on my own infrastructure?
Yes, you can self-host it using the open-source stack or deploy it on Azure, AWS, or in hybrid environments.
Is the platform vendor-locked?
No, it uses an adapter layer that allows you to swap different providers for storage, vector databases, and LLMs without changing code.