Zephex is a cloud-hosted MCP proxy that gives your coding agents access to 10 developer tools through a single API key at https://zephex.dev/mcp. No local Node servers, no per-editor maintenance — connect once and every tool call routes through one secure endpoint.
Start free with 555 requests per month. Typical sessions use 20–80 tool calls. Upgrade only when you need more volume.
Pick the path that matches how you work. Every route uses the same endpoint and API key — only the setup docs differ.
Editor user
Cursor, Claude Code, VS Code, Windsurf — one config, all 10 tools. Live in under 5 minutes.
Read guide →CLI / terminal
find_code, get_context, architecture — call MCP tools directly from the terminal without an agent.
Read guide →Team lead
One hosted endpoint, shared API keys, no per-laptop Node servers. Standardize configs across 20+ editors.
Read guide →Security reviewer
Supply-chain scans with check_package and project_memory. HTTP security audits with audit_headers.
Read guide →Your editor authenticates once with a Bearer token. The proxy routes each tool call to the right backend and returns a single response. Add or remove tools without ever touching your editor config.
Your editor sends a single JSON-RPC call. The proxy authenticates, routes to the right tool, returns one response.
Run this in your terminal. It walks you through API key creation and editor config automatically — no manual JSON editing required.
npx zephexWhy hosted MCP? Local MCP servers mean Node on every laptop, config drift across editors, and no central usage tracking. Zephex gives you get_project_context, read_code, find_code, check_package, project_memory, explain_architecture, Zephex_dev_info, check_test, audit_headers, and keep_thinking from one URL. See hosted vs local MCP.
The docs most people open after setup — setup walkthroughs, comparisons, and fixes for common issues.
Zero to first tool call in under 5 minutes.
Start hereLatest releases, new tools, and editor support.
What's newWhen a remote endpoint beats running Node on every laptop.
Session workflows that save tokens and avoid repeat mistakes.
Smaller tool outputs, smarter tool chains, lower bill.
0 tools showing? Auth errors? Step-by-step fixes.
1 request = 1 MCP tool call. No per-tool fees. No hidden backends charge. Start free, upgrade when your team needs more volume.
Free
$0
555 requests/mo · Community support
Pro
$7/mo
3,500 requests/mo · Email support
Max
$19/mo
10,000 requests/mo · Priority support
Follow the decision flow. Never guess files. Every tool page has copyable prompt templates and real Before/After examples.
Start here on any new task or unfamiliar codebase. Follow the arrows. Never guess files.
Every tool is available on every plan — including the free tier. Click any row for parameters, examples, and prompt templates.
Project context · code search · package audit · security · reasoning
| Tool | Category | Description | Access |
|---|---|---|---|
| get_project_context | Project | Detects framework, package manager, scripts, env vars | Read |
| check_test | Verification | Test Pulse — health checks, diagnostics, and test verification | ReadCompute |
| read_code | Code | AST-based symbol/file/outline reading | Read |
| find_code | Code | BM25-ranked code search across the repo | Read |
| explain_architecture | Code | End-to-end architecture analysis with Mermaid | ReadCompute |
| check_package | Packages | Live registry safety check (12 ecosystems) | Network |
| audit_headers | External | HTTP / TLS / cookie audit on any URL | Network |
| Zephex_dev_info | External | Expert dev knowledge base lookup | Read |
| keep_thinking | Reasoning | Persistent structured reasoning sessions | Compute |
| project_memory | Project | Remember and recall learned facts across sessions (local stdio) | ReadWrite |
See full details on the Capabilities page or any individual tool page.
Tools are grouped by job — start with project context, then drill into code, packages, or security as needed.
Trace how a repo fits together — stack, scripts, dependencies, entry points — in one call.
AST-based symbol reading, ranked code search, and architecture analysis with diagrams.
Check any npm, pypi, or cargo package for typosquats, CVEs, and supply-chain risks.
HTTP header audits, SSL checks, cookie flags, and CSP analysis on any deployed URL.
Persistent thinking sessions that prevent circular debugging and track conclusions.
Database schemas, security patterns, frontend/backend best practices, deployment guides.
Every method uses the same endpoint and API key. Pick the one that matches your workflow.
One command, auto-configures your editor. Works with Cursor, Claude Code, VS Code, and more.
→Full CLI with auth, key management, usage stats, and shell completions.
→find-code, get-context, architecture — run MCP tools in your shell (Mode 2).
→Direct JSON-RPC over HTTPS. Use from any language or custom integration.
→Create API keys from the dashboard. Each key gives access to all 10 tools through the proxy. Use separate keys per environment (dev/staging/prod) so you can rotate without downtime.
This config structure works in Cursor, Claude Code, VS Code, Windsurf, and any editor supporting MCP. Paste it, add your key, restart.
{ "mcpServers": { "zephex": { "command": "npx", "args": ["-y", "zephex"], "env": { "ZEPHEX_API_KEY": "YOUR_API_KEY" } } }}Same endpoint, same auth. What changes is where the config file lives.
Run this curl to confirm your key works and the endpoint is reachable.
curl -X POST https://zephex.dev/mcp \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'Tips for getting the most out of Zephex tools
→How the proxy routes requests to MCP backends
→Common issues and how to fix them
→Latest product updates and releases
→Rate limits, quotas, and how usage is tracked
→Encryption, data handling, and compliance
→