Do I need to install anything locally to use Zephex?
No. Zephex is cloud-hosted. The only local change is the MCP config you add in your editor. You do not need npm packages, local Node processes, or bridge servers just to connect.
Support
The fastest answers to setup, billing, security, and tool questions — grouped so you can scan by topic.
Plans: Free (555 req/mo), Pro (3,500 req/mo), Max (10,000 req/mo). Endpoint: https://zephex.dev/mcp · 10 tools. Not sure if Zephex matches your workflow? Start with who Zephex is for.
No. Zephex is cloud-hosted. The only local change is the MCP config you add in your editor. You do not need npm packages, local Node processes, or bridge servers just to connect.
Cursor, Claude Code, VS Code, Windsurf, JetBrains, and any other MCP-compatible client. If the editor supports the MCP spec, it works with Zephex. See the full list on /docs/supported-editors.
Most users connect in under 2 minutes — create an API key in the dashboard, add one config snippet to the editor, restart once. No npm install, no local servers.
Yes. Install Zephex from the VS Code Marketplace (works in VS Code, Cursor, and Windsurf), run Zephex: Sign In from the Command Palette, then enable Zephex MCP in agent mode. The extension handles OAuth and key storage — or run npx -y zephex@latest setup --vscode for stdio config. See /docs/editors/vscode-extension.
The web terminal at /dashboard/terminal runs Mode 2 mcpcli tools in your browser — paste a public GitHub repo URL and run overview, get-context, find-code, supply scans, and more without npm install. It uses a web-terminal-* API key and shares your plan quota. Private repos are not cloned server-side. See /docs/web-terminal.
Use one hosted endpoint (https://zephex.dev/mcp) with separate API keys per developer and per environment (dev/staging/prod). Do not commit keys to git. Standardize on npx -y zephex@latest setup globally, document editor restart requirements, and use dashboard usage analytics for quota visibility. Full checklist: /docs/team-rollout.
Yes, as long as the code is available on the machine you are working from. The code-analysis tools operate on project paths or file paths, not GitHub URLs. Private repos are not fetched automatically from GitHub by the hosted endpoint.
Local stdio servers run on your machine and are fast/private but create N×M config drift across editors, machines, and teammates. A hosted MCP gateway (remote HTTP) like Zephex gives every editor and every teammate the same 10 tools through one key and one endpoint. You get central usage analytics, key rotation, and production-grade tools (package CVE scanning, header audits, stateful keep_thinking) without shipping heavy scanners to every laptop. The tradeoff is one network hop and the rate limits of your plan.
Running MCP servers locally means managing Node processes, updating packages, and configuring each client separately. Zephex hosts the proxy for you — same endpoint, same access controls, no local infrastructure.
Free: 555 requests/month. Pro: 3,500 ($7/mo). Max: 10,000 ($19/mo). When you exceed your quota you receive HTTP 429 with a Retry-After header — share your link for +300 bonus requests or upgrade. Limits reset at the start of the next calendar month.
Free tier: 555 requests/mo. Pro ($7/mo): 3,500 requests. Max ($19/mo): 10,000 requests. Share your link for +300 bonus requests when someone clicks it.
Upgrade to a higher tier, wait for the monthly reset, or check npx -y zephex@latest usage. Rate limits are per API key and displayed in the dashboard. See /docs/plans for tier comparison.
Contact support within 14 days of payment for a refund on Pro or Max tiers. Free tier has no charges.
Yes. Contact support for custom tiers, SLAs, dedicated infrastructure, and custom integrations.
Code is processed ephemerally for the duration of the tool call only. Zephex does not train on customer code and does not retain source beyond what is required to answer the current request. See /docs/data-handling and /docs/support/security.
No. Zephex does not use customer prompts, tool inputs, tool outputs, or code for model training. See the Data Use page for the full policy.
Revoke it immediately from the dashboard and create a new one. Keys are shown once at creation — if you miss copying it, rotate to a fresh key. Then run npx -y zephex@latest reconnect --cursor (or your editor flag) to rewrite config.
Create a new key in Dashboard → API Keys, update your editor config (stdio env ZEPHEX_API_KEY or HTTP Bearer header), fully restart the editor, verify 10 tools appear, then revoke the old key. See /docs/api-keys for the full rotation sequence.
Yes. Create separate keys for local, staging, and production. Each key can have different backend access and shares your plan's monthly quota.
check_package is the fast first screen before any install — existence, typosquat signals, deprecation, and basic CVE/freshness across 12 registries (task=check|upgrade|security). project_memory lets your agent remember and recall project decisions, gotchas, and conventions across sessions via local SQLite — so you don't have to rediscover the same patterns every time. Rule of thumb: check_package before recommending a package; project_memory when you want your agent to persist context across sessions.
Google's managed MCP servers are excellent when your data already lives in BigQuery, Cloud Storage, or Firebase and you want IAM/OAuth + zero-ops. Zephex is the practical hosted gateway for general developer workflows across any stack: AST code reading, task scoping with risk assessment, real package migration + CVE intelligence across 12 ecosystems, HTTP header audits, and multi-step reasoning that actually persists state. It works with Cursor, Claude Code, Zed, JetBrains, Gemini CLI, and the full 19+ editor list from a single key. Use Google services for Google data; use Zephex for everything else an agent needs to understand and change real codebases.
We can help you get connected
Share your editor name, the error message (401, 429, empty tools/list), and your API key prefix — never the full key.