Give your AI agent its own place on the internet — a real public IPv6
address, a *.on.route6.me hostname, inbound port forwarding, and a private
encrypted mesh to reach your other agents. The agent manages all of it itself,
through 23 MCP tools. Free tier, no card.
- Website: https://route6.me · Docs: https://docs.route6.me · Pricing: https://route6.me/pricing
curl -fsSL https://dl.route6.me/install.sh | sh
mkdir -p ~/.r6me && chmod 700 ~/.r6me
echo 'api_key = "sk_a6_your_key"' > ~/.r6me/config.toml && chmod 600 ~/.r6me/config.toml
r6me up
r6me status # want "transport": "connected"Get the key from the dashboard under Agents.
r6me is one binary and runs as a normal user.
Other ways to run the same client:
| npm / pip | npm install -g @route6/agent or pip install route6, then route6 up — launchers that download and verify the same binary |
| Docker | docker/ — the same binary in an image |
| No client at all | point your MCP client at https://gw.route6.me/mcp with Authorization: Bearer <key>. Outbound tools only: no port forwards, no webhooks, no mesh |
The client serves MCP on http://localhost:3000/mcp, with no key in the
client config — the daemon already holds it. It also runs an egress proxy on
127.0.0.1:1080 (SOCKS5, HTTP CONNECT and plain HTTP), so your own code can
send traffic from the agent's address too.
Per-client setup: Claude Code · Codex · OpenCode · Cursor / Windsurf · Cline · Continue · LangChain / LangGraph · n8n · Hermes
| Example | What it shows |
|---|---|
| 01 — Public webhook URL | Receive Stripe/GitHub webhooks on a stable HTTPS URL, restricted to the sender's addresses |
| 02 — Your agent's own IP | Fetch from the agent's address, check its reputation, rotate when needed |
| 03 — Agent team coordination | Two agents share state and hand off work through a task queue |
| 04 — Agent-hosted service | Serve an API by name — publicly, or to your own agents only over the mesh |
Every example was run against the live service before it was published.
Every tool is available on every plan, including Free. Plans differ in capacity (number of agents, egress, scraper credits), not capability. A few things need a paid plan — choosing your own hostname, buying scraper credits, the Static IPv4 add-on, outbound SMTP — and the tool says so when you call it, naming the next step. Current numbers: route6.me/pricing.
Full tool reference: docs.route6.me/tools.
The examples in this repository are MIT. The Route6 client is proprietary: binaries are published at dl.route6.me, source is not.