NEDONION

rag-agent-platform: Self-Host RAG Agents and MCP Tools

A self-hosted Java and Next.js platform for combining document retrieval, LLM-powered agents, and MCP tools. It suits teams that want a configurable knowledge and agent workflow under their own infrastructure, with external services to operate.

ActiveJavaMITSelf HostedRAGMCP
View on GitHub Compare
Stars
89
Forks
7
Last commit
2 months ago
Contributors
1
Releases, 12 months
0

Our take

Activeregular commits

Health score

32/100

How it is scored
Activity
42
Community
4
Issues
21
Pull requests
62
  • Fewer contributors than 92% of the projects we track

This is a broad self-hosted RAG and agent console with active recent development, but its single-contributor concentration and infrastructure footprint make operational ownership important.

Good fit if

  • You want a single Java and Next.js application for knowledge bases, agents, model configuration, and MCP connections.
  • Your team can run PostgreSQL with PGVector, RabbitMQ, and S3-compatible storage, and is comfortable managing external model services.
  • You want the flexibility of the MIT license to run and modify the platform.

Look elsewhere if

  • You need a turnkey MCP tool-import experience, since the README says gateway and container setup is manual.
  • Your team cannot support the database, queue, object storage, and model-provider dependencies required for local deployment.
  • You need a project with a broader maintainer base or a recent release history. Recent commits and merged pull requests indicate activity, but most of the code comes from one person and there have been no releases in over a year.
All health signals
Last commit2026-08-18 (2 months ago)
Commits, last 90 days48
Releases, last 12 months0
Contributors1
Issues closed, last 90 days0
Pull requests merged, last 90 days4 (typically merged in 0 days)
Project age14 months

Checked on 2026-10-14 with the GitHub API.

Overview

RAG Agent Platform brings document ingestion, vector search, model configuration, agent conversations, and MCP tool connections into one self-hosted application. It addresses the need to manage knowledge-based LLM workflows while keeping control of the application and data infrastructure.

It is aimed at teams willing to run a Java backend, web frontend, database, message broker, and object storage. The README describes an incomplete tool-import workflow, so MCP gateways and tool containers need manual configuration.

Key Features

  • Upload and process documents for retrieval from knowledge bases.
  • Store vectors in PostgreSQL with the PGVector extension.
  • Configure model providers and publish agents.
  • Stream agent conversations to the frontend using Server-Sent Events.
  • Connect agents to a configured MCP gateway and tool containers.
  • Process document jobs asynchronously through RabbitMQ.
  • Deploy the frontend and backend with Docker Compose, with CI/CD deployment support documented in the repository.

Use Cases

  • A team can create an internal knowledge assistant over its documents when it can operate PostgreSQL, RabbitMQ, and object storage.
  • Platform engineers can provide configurable agents and model providers to users within a self-hosted workspace.
  • Developers can explore RAG and MCP integration in one application, especially when they are prepared to configure the MCP gateway and containers themselves.
  • Organizations that need to keep control of their deployment and data can evaluate the live demo first, then adapt the self-hosted setup to their infrastructure.

What you need

Detected in the repository

  • Java (JVM) (from pom.xml)
  • A Dockerfile, so it can run in a container; a Compose file sets up the related services
  • Automated checks on GitHub Actions

License in plain words

MITpermissive

  • Commercial use: yes
  • Modify and redistribute: yes
  • You must keep: the copyright and license notice
  • Share your changes: no

A summary, not legal advice: the LICENSE file is what applies.

Getting Started

Prepare Java 17, Maven, Node.js, pnpm, PostgreSQL with PGVector, RabbitMQ, object storage, and model-provider settings. Then start the backend and frontend:

mvn spring-boot:run
cd frontend && pnpm install && pnpm dev

See the README for database setup, environment variables, and deployment details.

Alternatives

  • 53AIHub: 53AIHub is an organization-facing portal that publishes and operates agents, prompts, tools, and knowledge bases, including integrations with other AI platforms.
  • rag-web-ui: RAG Web UI focuses on document ingestion and question answering, with configurable models and OpenAPI access rather than a broader agent and MCP platform.
  • MaxKB: MaxKB builds knowledge-grounded agents using workflows and external tools, while rag-agent-platform emphasizes configurable retrieval and MCP tool integration.
ProjectLanguageLicenseStarsStatus
rag-agent-platformJavaMIT89Active
53AIHubGoOther4.3kActive
rag-web-uiTypeScriptApache-2.03.3kQuiet
MaxKBPythonGPL-3.022.9kActive

Considerations

  • Deployment requires several services, including PostgreSQL with the vector extension, RabbitMQ, S3-compatible storage, model and reranker providers, and SMTP. The README does not list a GPU requirement.
  • MCP integration is not fully self-service in the UI. Teams must configure the gateway and tool containers manually.
  • The project is active, with regular recent commits and merged pull requests, but development is concentrated in one contributor. No releases were recorded in the past year, and no issues were closed in the past 90 days.
  • The supplied requirements data reports no tests. The README also cautions that its Compose setup is a deployment reference, not a one-command installation, and describes external infrastructure and private image-registry assumptions.
  • The MIT license permits broad use and modification, subject to its license terms and notice requirements.

Found this useful?

Share it with someone who would like rag-agent-platform.