Skillget

mcp-server-gateway

mcp
Security Audit
Pass
Health Pass
  • License — License: NOASSERTION
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 19 GitHub stars
Code Pass
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

MCP gateway & OAuth MCP servers. LangGraph agent: memory, eval/tracking, PII guard, context trimming. Docker quick start for enterprise agents.

README.md

MCP Server Gateway

License: Apache-2.0

Clone → Docker → working MCP stack. One repo for a Model Context Protocol gateway, MCP servers (Artifactory, GitHub, Google Drive), a LangGraph agent with memory, LLM tracking & eval, PII blocking, and context trimming—the pieces teams usually wire up separately.

Use it to learn governed MCP, fork it as a starter, or run the reference agent locally without shared API tokens in git.

Why developers use this

You get What it solves
MCP Gateway Single HTTP entry (:8090), sessions, optional per-user OAuth
MCP servers Artifactory, GitHub, GDrive—each behind the gateway
LangGraph agent Streamlit Command Center; tools via gateway; durable memory
Tracking & eval Run history, latency, cost estimates, quality views (SQLite local; BigQuery optional)
Agent Safety Kit PII redaction + secret block before the LLM; offline PII scan on stored text
Context trimming Token budgets and protocol-aware compression in the agent pipeline 
Developer / Cursor / LangGraph agent
              │
              ▼
     MCP Gateway  :8090
       OAuth · consent · proxy
              │
    ┌─────────┼─────────┐
    ▼         ▼         ▼
 Artifactory GitHub   GDrive
   MCP       MCP       MCP

Prerequisites

Tool Check
Docker + Compose docker compose version
Python 3.10+ For demo scripts & Agent Safety Kit (python3 --version)
Git Clone this repo

Optional later: Google/GitHub OAuth apps (secure mode), OpenAI/Anthropic key (smarter agent planner—not required).

Run in 10 minutes

1 — Gateway + MCP servers

Default compose starts the gateway, Redis, and backend MCP servers in open mode (no OAuth setup required to bring the stack up).

git clone https://github.com/harish-gaggar/mcp-server-gateway.git
cd mcp-server-gateway/mcp-gateway

cp .env.example .env
# Required: set TOKEN_ENCRYPTION_KEY in .env (paste output of next line)
openssl rand -base64 32

docker compose up -d --build
docker compose ps          # all services healthy
curl -s http://localhost:8090/health    # ok

Optional: point Artifactory at your instance in .env (ARTIFACTORY_BASE_URL, ARTIFACTORY_ACCESS_TOKEN). Without it, the Artifactory MCP process still runs for wiring tests.

Endpoints

URL Use
http://localhost:8090/health Gateway
http://localhost:8090/artifactory/mcp Artifactory via gateway
http://localhost:8091/mcp Artifactory direct (dev)

2 — LangGraph agent (memory, eval, context trim, PII guard)

With the gateway up on the host, start the reference agent (works without an LLM API key—offline planner included). Agent Safety Kit is baked into the image for pre-LLM guards.

cd ../Agents/jfrog-agent
cp .env.example .env
docker compose up --build

Open http://localhost:8501 — Command Center UI.

Area in UI Feature
Operations / Command LangGraph runs, MCP tools through gateway
Insights Tracking & eval (runs, tokens, latency)
Security / Governance Policy, approvals, redaction hooks
Agent pipeline Context trimming via jfrog_agent/context_optimizer/

Optional in .env: OPENAI_API_KEY or ANTHROPIC_API_KEY, JFROG_AGENT_LLM_PROVIDER=openai.

First-time Artifactory via gateway: complete the browser OAuth flow when the UI prompts you (see secure mode below if you enabled it).

3 — PII blocker framework (standalone, ~2 minutes)

No Docker required—use Agent Safety Kit in any agent (LangGraph, OpenAI loop, etc.).

cd ../../agent-safety-kit
python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest -v
python examples/01_guard_text.py
python examples/04_drop_into_any_agent.py

Details: agent-safety-kit/README.md.

Secure mode (per-user OAuth + Cursor)

For production-shaped auth—every user signs in with Google / GitHub, gateway consent, no shared tokens in the repo:

  1. In mcp-gateway/.env, uncomment and set:

    OAUTH_CONFIG_FILE=./configs/secure-oauth-config.yml
MCP_CONFIG_FILE=./configs/secure-mcp-config.yml
GATEWAY_BASE_URL=http://localhost:8090
GOOGLE_OAUTH_CLIENT_ID=...
GOOGLE_OAUTH_CLIENT_SECRET=...
GITHUB_OAUTH_CLIENT_ID=...
GITHUB_OAUTH_CLIENT_SECRET=...

  2. Follow doc/howto/QUICKSTART.md — create OAuth apps, wire Cursor, troubleshooting.

Smoke test without an IDE:

cd mcp-gateway
python3 scripts/oauth-client-demo.py --namespace github --tool list_repositories

Repository layout

mcp-server-gateway/
├── mcp-gateway/              # Gateway + docker compose + Cursor wrappers
├── artifactory/  github/  gdrive/   # MCP server implementations
├── agent-safety-kit/         # PII blocker + secrets guard (framework-agnostic)
├── Agents/jfrog-agent/       # LangGraph agent, memory, tracking, context trim
└── doc/howto/QUICKSTART.md   # Full OAuth + IDE setup

Documentation

Doc Contents
Quickstart (OAuth, Cursor, monitoring) Step-by-step secure stack
Gateway dev Config files, npm run dev, tests
JFrog agent Memory backends, telemetry, env reference
Context optimizer Trimming layers & presets
Contributing PRs and local tests

Project status

Active reference implementation—APIs may evolve. Issues, stars, and forks help; contributions welcome.

Security

Never commit .env or OAuth secrets. Prefer secure mode for shared environments. Run input guard on all user/tool text before LLM calls.

License

Apache-2.0

Reviews (0)

No results found