A community-driven registry for the Claude Code ecosystem. Not affiliated with Anthropic.
Are you the author? Sign in to claim
Production-grade MCP server for MikroTik RouterOS with secure AI-native network automation.
AI-native network automation for MikroTik RouterOS. MikroMCP exposes RouterOS as a typed, auditable Model Context Protocol server so Claude, Cursor, Codex, and other MCP clients can inspect, diagnose, and safely operate MikroTik routers in natural language.
MikroMCP exists because raw router CLI access is the wrong abstraction for AI agents. RouterOS is powerful, but asking an LLM to improvise shell commands against production network gear is risky. MikroMCP gives agents a controlled tool surface: strict schemas, idempotent writes, dry-run previews, per-router circuit breakers, retry policies, RBAC, audit logs, snapshots, and rollback-aware change workflows.
In one sentence: MikroMCP turns MikroTik RouterOS into a production-minded MCP control plane for AI infrastructure, DevOps automation, and modern router management.
That's the whole setup for a single-router stdio deployment. For standalone binaries, Docker, HTTP/SSE mode, the RouterOS API prerequisites, and the full 15-minute walkthrough, see the Getting Started guide.
| Category | What MikroMCP covers |
|---|---|
| 🧭 Router management | System status, clock, reboot, packages, files, scripts, scheduler jobs, containers |
| 🌐 Network operations | Interfaces, VLANs, IP addresses, DHCP leases, DNS static records, bridge ports, WiFi clients |
| 🔥 Firewall and policy | Filter/NAT rules, mangle rules, address lists, route tables, routing rules |
| 🛰️ Routing visibility | Static routes, routing tables, BGP peers, OSPF neighbors |
| 🔐 Secure access | HTTP bearer auth, bcrypt token hashes, RBAC, router/tool restrictions, confirmation tokens |
| 🧪 Diagnostics | Router-originated ping, traceroute, torch, log filtering, guarded SSH command execution |
| 🛡️ Change safety | Dry-run, idempotent writes, snapshots, write journal, plan_changes, apply_plan, rollback_change |
| ⚙️ Production behavior | Retries for read tools, per-router circuit breakers, correlation IDs, structured logs, audit logs |
| 🤖 AI-agent fit | Human-readable responses plus structured JSON content for reasoning, chaining, and automation; server advertises an instructions string on MCP initialize so clients self-configure; optional routerId resolved via MIKROMCP_DEFAULT_ROUTER for single-router setups; usage skill for safe, guided tool use in Claude Code |
| 🧩 MCP compatibility | stdio for desktop clients, Streamable HTTP and legacy SSE for remote or service-style clients |
117 typed tools in total — browse the full catalog with parameters, defaults, and copy-paste example prompts in Available Tools.
Use MikroMCP to inspect core-01. Summarize system resources, RouterOS version,
running interfaces, active routes, DNS settings, and recent warning/error logs.
Flag anything that looks operationally risky.
List firewall filter and NAT rules on edge-01. Identify disabled rules,
overlapping port forwards, broad accept rules, and anything without comments.
Do not change anything yet.
Dry-run a route on core-01 for 10.20.0.0/16 via 192.168.88.1 in the main table.
Show the exact planned diff and tell me whether an existing route conflicts.
Show WireGuard peers on branch-02. Sort by last handshake age and flag peers
that have not handshaken recently or have no transfer counters.
Check interface health on edge-01, then run ping and traceroute from the router
to 1.1.1.1. If packet loss is present, use torch on the WAN interface for a
short traffic snapshot.
Create a change plan that adds a DNS record and a firewall address-list entry
on edge-01. Use dry-run first, explain the plan, then wait for approval before
applying anything.
MCP gives LLMs a standard way to call tools. MikroMCP makes RouterOS a high-quality MCP target by turning network operations into well-described, machine-readable, permission-aware actions.
AI assistants can use MikroMCP to:
MikroMCP is an open-source Model Context Protocol (MCP) server that exposes MikroTik RouterOS as 117 typed, auditable tools — letting AI assistants inspect, diagnose, and safely operate routers in natural language instead of improvising CLI commands.
The RouterOS REST/API exposes raw endpoints. MikroMCP wraps them in schema-validated, idempotent, dry-run-able tools with RBAC, audit logging, snapshots, and rollback — the safety layer an LLM needs before it touches production gear.
Instead of brittle SSH scripts that screen-scrape CLI output, MikroMCP returns structured, typed results with confirmation gates and per-router circuit breakers. SSH is used only where REST can't reach — ping, traceroute, torch, and guarded run_command.
MikroMCP speaks MCP over stdio and HTTP/SSE, so Claude Code and Claude Desktop drive RouterOS directly. Pair it with the bundled usage skill for safe, guided workflows.
Codex connects to MikroMCP over the standard MCP protocol — see Connecting to AI Assistants.
Cursor connects to MikroMCP as an MCP server (stdio or HTTP) to inspect and manage MikroTik routers without leaving the editor.
Any MCP-compatible client — OpenClaw included — can use MikroMCP; configure it as a stdio or HTTP MCP server.
Start with Getting Started to install and connect, then use the usage skill and Available Tools to automate RouterOS safely with an AI assistant.
MikroMCP is purpose-built for MikroTik/RouterOS operations with production-grade safety — dry-run, rollback, audit, and RBAC — making it a strong MCP choice for network engineers adopting AI tooling.
The README stays intentionally short. Everything below is documented in depth in the wiki:
| Resource | Use it for |
|---|---|
| Getting Started | Install (npm, binary, Docker), configure, and connect in 15 minutes |
| RouterOS API Setup | Enable the REST API, create a user, TLS and firewall |
| Configuration | Router registry, credentials, all environment variables |
| Running | Run commands, HTTP/SSE transport, troubleshooting |
| Connecting to Claude Desktop | Register MikroMCP in Claude Desktop |
| Connecting to AI Assistants | Claude Code, Cursor, Codex, HTTP/Docker/systemd |
| Using the Skill | Install the MikroMCP usage skill so your assistant drives the tools safely |
| Available Tools | All 117 tools — parameters and example prompts |
| Architecture | System layers, request pipeline, auth model |
| Error Handling | Error categories, retry engine, circuit breaker |
| Security | Threat model, hardening checklist, vulnerability reporting |
| Development | Project structure, tests, MCP Inspector workflow |
| Contributing | Adding tools, coding conventions, PR checklist |
| Roadmap · ROADMAP.md | Shipped milestones and guiding principles |
Issues, bug reports, tool requests, documentation improvements, and pull requests are welcome.
Good first contributions:
Development standards:
.js extensions.strict(), idempotency and dryRun for write toolsMikroMCPError for domain errors, focused Vitest coverage for every toolPlease open an issue before large changes so maintainers can align on scope.
MikroMCP controls real network devices — treat it like an operations system: least-privilege RouterOS users, verified TLS (or pinned fingerprints), credentials only in ~/.mikromcp/.env, scoped RBAC identities, and audit logging for shared use. The full hardening checklist and vulnerability-reporting process are on the Security page.
MikroMCP is released under the MIT License.
Run Claude Code as an MCP server so any agent can delegate coding tasks to it
Browser automation using accessibility snapshots instead of screenshots
@designcomputerSecure MCP server for MySQL database interaction, queries, and schema management
English-first Korean equity intelligence MCP — DART filings, foreign-holder 5%-rule flows, activist filings, KRX news. F
