mcpField
M
mnki-mcp

mnki-mcp

Local
mnki.comSecurity1Updated 5w ago
RepositoryVisit website

About mnki-mcp

mnki-mcp is a local proxy for MCP clients and servers that checks each tool call against Agent Trust before forwarding it. Developers can choose logging, warnings, approval escalation, or enforcement, with evidence surfaced for decisions.

Key features

  1. 01Intercept and verify tool calls
  2. 02Observe, warn, approve, or enforce modes
  3. 03Forward stdio and Streamable HTTP servers
  4. 04Provide evidence with decisions

Use cases

  1. 01Log tool-call checks without blocking
  2. 02Show policy evidence while forwarding calls
  3. 03Escalate selected calls for human approval
  4. 04Deny calls under enforcement policy

MCP support

The MCP proxy sits between a client and an MCP server, forwarding traffic while checking every `tools/call` against Agent Trust. It supports stdio commands and Streamable HTTP endpoints.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
  "mcpServers": {
    "mnkiagentos-mnki": {
      "command": "npx",
      "args": [
        "-y",
        "mnki-mcp"
      ],
      "env": {
        "MNKI_API_KEY": "<MNKI_API_KEY>",
        "MNKI_URL": "<MNKI_URL>"
      }
    }
  }
}
Run commandnpx -y mnki-mcp
Environment variables to set
MNKI_API_KEYSecretAgent Trust API key with the verify scope (console → Settings → API keys).
MNKI_URLConsole base URL; https://mnki.com unless self-hosted.

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmmnki-mcp0.3.0

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

mnki-mcp

mnki-mcp is a local MCP proxy: it sits between an MCP client (Claude Desktop, Claude Code, Cursor, Windsurf, VS Code) and an MCP server — a stdio command or a Streamable HTTP endpoint — and verifies every tools/call against Agent Trust before it reaches the server. Everything else is forwarded untouched.

Modes (--mode): observe (log only, always forward — the default), warn (forward, surface evidence), require_approval (escalate only where policy says so, never hard-deny), enforce (deny with evidence, wait for human approval).

JSON-RPC errors the client sees in enforce mode: -32003 denied (with the ✓/⚠/✕ evidence), -32001 approval required / rejected / expired, -32005 verification quota exceeded, -32006 Agent Trust unreachable (observe and warn forward instead).

Install it with mnki protect (dry run by default) or by hand — see examples/.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.