FAQ
Short answers to what people ask before trying runelay.
Doesn't a normal agent harness already do this?
Each harness does it for itself. Claude Code, Codex, Cursor, Grok, Gemini and OpenCode each keep their own MCP config, their own logins, their own memory and their own skills folder, per tool and per machine. Add Linear to one and the others never see it; teach one a rule and the others never hear it.
runelay sits under all of them. You connect an account once, on your server, and every agent gets the same tools through one MCP endpoint. Memories and skills live in one place too, so what one agent writes, the others read. Adding a new agent is one key and one line of config, not a fresh round of logins.
It is self-hosted: one container, or the desktop app, on hardware you control.
Is my data private?
runelay runs on your server or your computer. There is no runelay cloud and no third-party service in between: your agents talk to your runelay, and runelay talks directly to the accounts you connected.
Tokens and connection configs stay on your server, encrypted with AES-256-GCM using your RUNELAY_SECRET. Agents never receive them: they call tools through runelay and see credentials masked as ••••••. See Security.
How much does it cost?
Nothing. runelay is free and open source under the MIT license. If it saves you time, you can support it on GitHub Sponsors.
Does runelay run a model?
No. runelay runs no model and calls no LLM. Your agent does the thinking with whatever model it already uses; runelay is where it finds tools, memories and skills.
Which agents work with it?
Any agent that speaks MCP over HTTP with a bearer token. runelay connect wires Claude Code, Codex, Cursor and Grok for you, and runelay migrate also imports from Gemini and OpenCode. See Connect agents and Migrate from your agents.
What if something breaks?
See Troubleshooting.