ProtoMCP
Health Uyari
- No license — Repository has no license file
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Community trust — 13 GitHub stars
Code Gecti
- Code scan — Scanned 2 files during light audit, no dangerous patterns found
Permissions Gecti
- Permissions — No dangerous permissions requested
Bu listing icin henuz AI raporu yok.
A Postman-style lab for exploring, testing and debugging Model Context Protocol (MCP) servers
ProtoMCP
Postman for MCP servers — connect, explore, and test from your browser
Live Demo · Docs · Get Started
ProtoMCP is a browser-based inspector for MCP servers. Point it at any MCP server — local or remote — and you get a full UI to browse tools, prompts, and resources, fire requests, and watch every JSON-RPC exchange in real time.
Why ProtoMCP?
- Browser-only — No native app, no extensions, no localhost stdio gaps
- Multi-server — Connect 13+ public MCP servers in a single workspace
- Real-time trace — Watch every JSON-RPC exchange live (Claude Desktop doesn't show this)
- Zero-config introspection — Auto-discovers tools, prompts, resources from schemas
How It Compares
| ProtoMCP | Claude Desktop | MCP Inspector (CLI) | |
|---|---|---|---|
| Browser-based | ✅ | ❌ | ❌ |
| Multi-server | ✅ | ✅ | ❌ |
| JSON-RPC trace | ✅ | ❌ | ✅ |
| Auto-generated forms | ✅ | ❌ | ❌ |
| Agent mode | ✅ | ✅ | ❌ |
| Choose LLM provider | ✅ | ❌ | ❌ |
| No install needed | ✅ | ❌ | ❌ |
What it does
Explore Mode
Connects to any MCP server over Streamable HTTP or SSE. Once connected, ProtoMCP discovers all capabilities automatically and generates input forms from the server's JSON Schema — so you can invoke tools, preview prompts, and browse resources without writing a single line of code. Every request and response is logged with timing data in a live panel alongside.
Agent Mode
Drops an AI assistant on top of your connected servers. Pick a provider (OpenAI, Anthropic, Gemini, Groq, or others), write a prompt, and watch the agent reason and call tools across all connected servers in real time. You can require confirmation before each tool call if you want to stay in the loop — useful when the tools have side effects.
Multiple servers can be connected at the same time. Tools and resources from all of them are aggregated into a single workspace, so you can mix and match across servers in a single agent session.
Get Started
Hosted — no setup, open in your browser:
Run locally:
pip install jaclang==0.12.0 jac-client==1.0.7 byllm jac-mcp jasketch-mcp-server
git clone https://github.com/SahanUday/ProtoMCP.git && cd ProtoMCP
jac start main.jac
Open http://localhost:8000. To connect a local MCP server, see the ngrok tunnel guide.
Featured Servers
| Name | Transport | Use Case |
|---|---|---|
| GitHub Copilot | HTTP | Code-aware search |
| Exa Search | HTTP | AI web search |
| Hugging Face | HTTP | Model metadata |
Built-in servers (no URL required):
jac-mcp— Jac language validation, formatting, and docsjasketch— Live diagram canvas
Transport Support
| Transport | Description | Session Management |
|---|---|---|
| Streamable HTTP | Modern, session-aware | Auto-managed via mcp-session-id header |
| SSE | Legacy Server-Sent Events | Connection-based, auto-discovered |
Authentication
| Type | Header Config |
|---|---|
| Bearer Token | Authorization: Bearer <token> |
| API Key | Custom header name + value |
| Basic Auth | Authorization: Basic <base64> |
| None | No headers |
Architecture
graph TB
User(["User"])
subgraph Browser["Browser"]
UI["Explore / Agent UI"]
end
subgraph Backend["Jac Backend"]
Proxy["CORS Proxy"]
Agent["Agent Runner"]
end
subgraph External["External Services"]
MCP["MCP Servers"]
LLM["LLM Providers"]
end
User --> UI
UI -->|"agent requests"| Agent
Proxy <-->|"JSON-RPC"| MCP
Agent <-->|"tool use"| Proxy
Agent <-->|"chat"| LLM
Agent -->|"streaming trace"| UI
The Jac backend bypasses browser CORS restrictions by proxying all MCP requests. It pools SSE connections for legacy servers and runs the ReAct loop for Agent Mode, streaming the full trace back to the client in real time.
Tech Stack
| Layer | Technology |
|---|---|
| Framework | Jac (Jaseci) v0.12.0+ |
| Frontend | jac-client v1.0.7 |
| Styling | Tailwind CSS v4 |
| Editor | Monaco Editor |
| LLM routing | byllm |
| Deployment | Kubernetes via jac-scale |
Configuration
Environment Variables
| Variable | Description | Default |
|---|---|---|
JAC_SCALE_NAMESPACE |
K8s namespace for deployment | jac-mcp-playground |
Session Storage
mcp_saved_connections— Connection configs (URL, auth, transport)mcp_agent_session_history— Agent chat sessions (max 5)
Roadmap
- Desktop App — Native app via Tauri with direct localhost and stdio transport support. See the Desktop App docs.
Development
# Install dependencies
pip install jaclang jac-client byllm jac-mcp jasketch-mcp-server
# Start dev server
jac start main.jac
# Run with hot reload (file watching)
pip install watchdog
jac start main.jac --watch
Adding a New Server to Registry
Edit data/mcp_servers.cl.jac:
{
"id": "your-server",
"name": "Your Server",
"namespace": "example.com",
"transport": "http", # or "sse"
"description": "Single-line description",
"url": "https://example.com/mcp",
"auth": "bearer", # or "apikey", "basic", "none"
"authRequired": True
}
Contributing
Issues and pull requests are welcome. If you're fixing a bug, open a PR directly. If you're adding something new, open an issue first so we can align on the approach before you build it.
Check open issues for things tagged good first issue or help wanted.
License
MIT — see LICENSE for details.
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi