mcp-gatewayBLOCK
🔀 A lightweight, open-source gateway for managing, routing, and monitoring multiple MCP (Model Context Protocol) servers. Auth · Rate-limit · Metrics · Tool discovery.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
[](https://www.npmjs.com/package/@winstonsayno/mcp-gateway) [](https://github.com/HarrisonCN/mcp-gateway/actions/workflows/ci.yml) [](https://github.com/HarrisonCN/mcp-gateway/actions/workflows/codeql.yml) [](https://securityscorecards.dev/viewer/?uri=github.com/HarrisonCN/mcp-gateway) [](https://nodejs.org) [](LICENSE)
English · 简体中文
One authenticated, observable endpoint in front of all your MCP servers.
mcp-gateway sits between AI clients (Claude Code, Cursor, your own agents, LLM apps, web and mobile front-ends) and the Model Context Protocol servers they use. Clients connect once — over MCP Streamable HTTP at /mcp or a REST API under /api/v1 — and the gateway authenticates the caller, checks what it may do, routes the call to the right upstream and records what happened. Keys, scopes, rate limits, policy, logs and metrics live in one place instead of in every client.
Current release: 13.6.0 (npm latest). Live dashboard demo with simulated traffic, no backend:
Contents
- Architecture
- Key features
- Quick start
- Minimal configuration
- Connect clients and LLMs
- Security model
- Observability
- Supported versions
- Documentation
- Contributing
- [Licens
ca6da521050aOBSERVED · 2026-10-10Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add mcp-gateway --env AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID} --env AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY} --env AWS_SESSION_TOKEN=${AWS_SESSION_TOKEN} --env CI_KEY=${CI_KEY} -- npx -y @winstonsayno/[email protected]{
"mcpServers": {
"mcp-gateway": {
"command": "npx",
"args": [
"-y",
"@winstonsayno/[email protected]"
],
"env": {
"AWS_ACCESS_KEY_ID": "${AWS_ACCESS_KEY_ID}",
"AWS_SECRET_ACCESS_KEY": "${AWS_SECRET_ACCESS_KEY}",
"AWS_SESSION_TOKEN": "${AWS_SESSION_TOKEN}",
"CI_KEY": "${CI_KEY}"
}
}
}
}Exposed tools (39)
32 read · 6 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
Extra | read | d |
Fake | read | test |
Fetch | read | Fetch web pages and convert them to markdown |
Filesystem | write | Read and write files under /srv/projects |
GITHUB_TOKEN | read | GitHub personal access token |
Git | read | Read, search and manipulate a Git repository |
GitHub | read | Repositories, issues and pull requests |
Memory | read | Knowledge-graph based persistent memory |
Postgres | read | Read-only SQL against the analytics replica |
Slack | write | Post and read team messages |
Time | read | Current time and time-zone conversion |
a__echo | read | |
a__n | read | d |
add-tool | write | |
brave_local_search | read | Search for local businesses and places |
brave_web_search | read | Search the web |
cached | read | Cacheable echo |
create_issue | write | Open a new issue |
directory | read | Allowed directory |
echo | read | Echo the arguments |
fail | read | Always a JSON-RPC error |
get_issue | read | Get an issue by number |
late | read | |
list_directory | read | List files and folders in a directory |
list_pull_requests | read | List pull requests in a repository |
list_tables | read | List tables in the public schema |
priced | read | Echo with a usage block (budgets) |
query | write | Run a read-only SQL query |
read | read | Read |
read_file | read | Read the complete contents of a file |
repository | read | Path of the repository |
search_files | read | Recursively search for files matching a pattern |
search_repositories | read | Search GitHub repositories |
slack_get_channel_history | read | Recent messages in a channel |
slack_post_message | write | Post a message to a channel |
slow | read | Sleep |
twice | read | echo twice |
write_file | destructive | Create or overwrite a file |
x | read |
Trust audit
BLOCKgrade F · trust 39/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (8 observation(s))
- Network
- declared (13 observation(s))
- Shell
- declared (6 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- found
Findings (25)
} else socket.write(exec(v));
st.done = executeRun(run, exec(run), { checkpoint: (r) => taskRuns.save(r, dirOf(c)), cancelled: () => st.cancel }).finally(() => active.delete(run.id));exec(sql: string): void;
eval(e: Expr): CedarValue {# Opt out of the auth requirement: start the gateway with --insecure (NO authentication on the pod address).
.option('--insecure', 'Allow starting without auth on a non-loopback address (trusted networks only)')const s = connect({ host, port, servername: /^[\d.]+$|:/.test(servername) ? undefined : servername, ecdhCurve, rejectUnauthorized: false, minVersion: 'TLSv1.3' }, () => {if (this.config.security?.insecure && (this.config.auth?.strategy ?? 'none') === 'none') logger.warn('Started with --insecure: authentication is off on a non-loopback address.');'127.0.0.1, or — only on a trusted network — start with --insecure (security.insecure: true).'
gradle-wrapper.jar
const nr = { id: uuid(), timestamp: Date.now(), serverId: s.id, toolName: r.toolName, durationMs: Math.round(s.base * (0.5 + Math.random())), success: up, clientId: 'key:demo', via: 'rest', kind: 'toologger.warn(`OAuth token introspection failed: ${err instanceof Error ? err.message : String(err)}`);logger.warn(`OAuth token rejected from ${req.ip}: ${te.message}`);console.log(`Wrote ${options.out}.key (keep secret) and ${options.out}.pub (add to pluginTrust.keys)`);console.log(`Wrote ${options.out}.hybrid.json (secret) and ${options.out}.pub.json (Ed25519 + ${options.alg}). Backend: ${pqBackend().note}`);logger.info(`Secret rotation requested via API: ${rotated.length} server(s) reconnected`);{ id: "sanitize", since: "7.3.0", summary: "Prompt-injection defence: tool-output sanitisation (hidden Unicode, ANSI, HTML, exfil images), spotlighting, inbound / outbound blocking" },{ timestamp: iso(Date.now() - 64e3), client: 'key:aura', kind: 'prompt-injection', detail: 'results: ignore-instructions, exfil-url', serverId: 'web', tool: 'fetch', score: 1 },if (p === '/admin/anomaly/score' && method === 'POST') return json({ score: 1, signals: ['ignore-instructions', 'prompt-exfiltration'], threshold: 0.6 });'prompt-exfiltration': 0.5,
'exfil-url': 0.7,
'key sk-ant-REDACTEDREDACTEDREDACTED00',
'aws AKIAABCDEFGHIJKLMNOP',
const SECRET = 'federation-secret-0123456789abcdef';
const input = { serverId: 's', tool: 't', args: { token: 'sk-abcdefghijklmnopqrstuvwxyz123456' }, via: 'rest' as const, clientId: 'key:app' };Gates applied: no_behavioural_pass.
ca6da521050afull audit observations/trust-audit/mcp-server/harrisoncn__mcp-gateway.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-10 | ca6da521050a | BLOCK | F | 39 | first audit |
Questions
What is the @winstonsayno/mcp-gateway MCP server?
🔀 A lightweight, open-source gateway for managing, routing, and monitoring multiple MCP (Model Context Protocol) servers. Auth · Rate-limit · Metrics · Tool discovery.
What tools does @winstonsayno/mcp-gateway expose?
39 in total: 32 read-only, 6 that write, and 1 that can delete or overwrite (write_file). Every one is listed on this page with its risk.
Is @winstonsayno/mcp-gateway safe to connect to an agent?
No — not without reading the findings first. The audit graded it F (39/100) and found 9 critical or high issues in the source. Each one is listed on this page with the file and line it is on. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does @winstonsayno/mcp-gateway need?
It reads AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN, CI_KEY, GATEWAY_KEY, GH_TOKEN, GITHUB_TOKEN, GOOGLE_OAUTH_ACCESS_TOKEN, LLM_API_KEY, MCP_GATEWAY_ADMIN_KEY, MCP_GATEWAY_API_KEYS and MCP_GATEWAY_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does @winstonsayno/mcp-gateway run?
It speaks sse and streamable-http, so it runs as a service you connect to over the network. It is published on npm as @winstonsayno/mcp-gateway at 13.6.0.
How current is this page?
The grade is for one exact copy of the source (ca6da521050a), read on 2026-10-10. The repository is watched and re-audited when it changes.