Product
Why BeyDock lets agents hand work to each other
A session stays on one provider, but its work doesn't have to. Delegation lets an agent start child sessions on other providers and accounts, then read their results.
Notes from the dock.
What's new in BeyDock, how we build it, and how to get more out of your agents.
Product
Workflows: agent work that starts itself
A trigger starts one or more agent sessions on the provider and account you pick. Build the flow on a canvas, review the result, and approve anything that touches the outside world.
Engineering
Keeping a Mac awake with the lid closed
Sleep assertions do not stop lid sleep. Here is how a root helper, an XPC connection and a long list of undo paths let agents keep working when you close the laptop.
Engineering
The BeyDock Hub: one MCP setup for every agent
How we connect third-party MCP servers once and use them from every provider and account, with tokens kept in the Keychain, per-tool policies and an activity log.
Product
Permissions you can read
BeyDock's approval cards show the full command and every changed file with its diff, and Allow never approves more than what is on screen. Here are the rules behind that.
Engineering
Sessions that outlive your laptop: agents on SSH hosts
How BeyDock runs agents on a remote machine with a long-lived host daemon, a sequence-numbered replay after reconnects, an automatic sidecar deploy and a tunnel back to the Hub.
Engineering
One app, four agent protocols
How the BeyDock sidecar routes commands to Claude, Codex, Antigravity and OpenCode, translates their events into one shape, and switches off what a provider cannot do.
Guides
Getting set up: accounts, limits, SSH and worktrees
A walkthrough of the first things to configure in BeyDock: add accounts for each provider, choose the model and effort per session, read your usage limits, connect an SSH host and isolate work in worktrees.