Why we built vemlor
A control plane for coding agents, so you bring the agent and the subscription, and we run everything around it.
Coding agents got good. The plumbing around them did not.
Every team that wants an agent working the merge queue ends up rebuilding the same scaffolding: a runner to execute the agent, auth to the right repos, webhooks to know when there is work, and an audit trail so you can trust what happened. That is the boring part, and it is the part that keeps agents stuck in a terminal instead of shipping.
Bring your own agent
vemlor is the control plane, not the model. You point it at Claude Code, Codex, or whatever ships next, using your own subscription. We run the runners, the integrations, the auth, the events, and the audit. GitLab and GitHub are first-class peers, not an afterthought.
What this looks like
An assignment, a comment, a webhook, or a chat message becomes a task. The platform picks it up and takes it to a reviewed merge request. You make the calls; the agent does the typing.
This is the first post. More to come on how the runners work, how we keep agent runs auditable, and what we have learned running this in production.