Introduction
Rilo is GTM infra for AI agents - connect your agent once and it runs your LinkedIn outbound.
Rilo is an MCP server. Point your AI agent at it once and the agent can find leads, qualify them, build sequences, and send on LinkedIn as you - while Rilo handles the mechanical work: provider integrations, storage, scheduling, and safety limits.
Your agent gets ~50 tools across seven areas: search & enrichment, sequences, LinkedIn actions, inbox, learning, connect, and observability. The subjective judgment - who to target, what to say - stays with your agent. Rilo is the hands, not the brain.
Connect your agent
For terminal-based agents, one command writes the Rilo configuration:
npx rilogtm connect
Detects your agents and adds the Rilo endpoint. The client handles sign-in afterward when its remote MCP support allows it.
Or pick your client for a step-by-step guide:
claude.ai / Claude Desktop
Add Rilo as a custom connector. No terminal, no key.
ChatGPT
Connector OAuth is not production-verified yet.
Claude Code
One command in your terminal.
Cursor
One command writes the MCP config.
Codex
One command in your terminal.
OpenClaw
One command writes the config for you.
HTTP API
Same operations over plain HTTP, for your own code.
CLI
Wire up any client from your shell.
One response shape everywhere
Every operation - over MCP, HTTP, or the CLI - returns the same envelope:
{
"ok": true,
"data": {},
"summary": "…",
"signals": [],
"learnings": []
}summary says what happened in one line and signals flags anything unusual
worth the agent's attention. Details in the reference.
Where to go next
- Quickstart - connect your client and send a first task.
- Concepts - lead groups, sequences, inbox, knowledge, observability.
- Tools - every tool, with parameters and sample requests.