A
licenseNot graded
qualityB
maintenanceEnables document-based Q&A with multi-modal RAG, hybrid retrieval, knowledge graph reasoning, and multi-agent orchestration via MCP tools.
4
MIT
mcp-rag-platform
by mittal-umang
Allows the RAG agent to use OpenAI language models as the LLM provider for generating grounded, retrieval-augmented responses.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-rag-platformRetrieve sources on ArgoCD rollback and include citations."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mcp-rag-platform
A production-style RAG agent served behind an MCP server, wrapped in the platform layer that makes it operable: containerized, GitOps-deployed on Kubernetes (Helm + ArgoCD), and observable. The agent is the payload; the serving/reliability platform around it is the point.
Personal project. Not affiliated with any employer.
Architecture
┌─────────────────────────────────────────────┐
Claude Desktop ───▶│ MCP server (retrieve / fetch_source / │
(MCP host) │ ingest) │
└───────────────┬─────────────────────────────┘
│
┌───────────────▼─────────────────────────────┐
HTTP ────────────▶│ Agent (FastAPI + Pydantic) │
/query │ embed → retrieve → ground → generate │
│ weak-retrieval guardrail │
└───────┬───────────────────────┬─────────────┘
│ │
┌────────▼─────────┐ ┌────────▼──────────────┐
│ Qdrant (vectors) │ │ LLM provider │
└──────────────────┘ │ anthropic | openai │
│ | vllm (Phase 2) │
embeddings: local └───────────────────────┘
(sentence-transformers)Related MCP server: MCP RAG Server
Quickstart (local)
cp .env.example .env # add your provider API key
make up # docker compose: agent + qdrant
make index # embed + upsert the corpus into qdrant
make ask Q="how do I roll back an argocd app?"Deploy to a local cluster (kind + Helm + ArgoCD)
make kind-up # create kind cluster
make deploy # helm install the chart
make argocd # apply the ArgoCD Application (GitOps)
make port-forward # expose the agent locallyMCP tools
tool | purpose |
| top-k grounded chunks with source citations |
| full document/section behind a chunk |
| add/re-index a document (idempotent per doc_id) |
Observability
GET /metrics exposes Prometheus counters/histograms (request count, retrieval latency,
generation latency, tokens). Structured JSON logs on stdout. See docs note in
src/common/metrics.py for wiring Datadog.
Phase 2 - self-hosted inference (vLLM)
The generation layer is provider-pluggable. Set LLM_PROVIDER=vllm and enable the gated
vLLM service (vllm.enabled=true in Helm values) to serve a small open model from an
OpenAI-compatible endpoint, with continuous batching and token streaming. Needs a GPU;
the API providers remain the zero-GPU default. See src/agent/providers/vllm_provider.py.
Layout
src/agent FastAPI app, RAG pipeline, pluggable LLM providers
src/mcp_server MCP tools (retrieve, fetch_source, ingest)
src/indexing loaders, chunking, embeddings, qdrant store
src/common pydantic models, config, logging, metrics
deploy/helm chart: agent, mcp-server, qdrant, (vllm)
deploy/argocd Application manifest (GitOps)
deploy/kind local cluster config + bootstrap
data corpus manifest
tests unit tests + eval harnessThis server cannot be installed
Maintenance
ActivityMaintained
ResponsivenessNo issues
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Connectors
- docs2mcpOAuthcom.docs2mcp
Query your own PDFs and documents from any MCP client. Every answer cites the page it came from.
Ingest, manage, and retrieve documents for RAG-powered AI applications
Query any docs site via MCP. Submit a URL, ask questions, get cited answers.
Team docs served to AI agents over MCP - search, Markdown reads, version pinning, read audit.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables document ingestion, semantic search, and retrieval-augmented generation via MCP tools and REST API, using vector embeddings and intelligent chunking.MIT
- FlicenseNot gradedqualityCmaintenanceProvides read-only, citation-backed semantic search and retrieval-augmented generation over enterprise documents via standardized MCP tools, with local embeddings for privacy.-
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to query and manage a document knowledge base via MCP, with RAG-powered search and grounded answers with citations.MIT