Renovate did dependency updates.
vemlor does everything else.
Bot opens an MR, you review, you merge.
Same loop, but the bot fixes failing pipelines.
Built for self-hosted GitLab teams in regulated EU.
Built on the agents you already pay for:
How vemlor works
Three moving pieces. Same loop as Renovate, but the bot is an agent. Drop it onto a self-hosted GitLab today.
POST /webhook/gitlab
{
"event": "note",
"comment": "@vemlor fix the flake"
}Webhook fires
Push, MR open, comment, or CI fail. Your GitLab pings vemlor, we sign the event, queue a task.
- ✓ workspace acquired
- ✓ repo cloned (warm cache)
- claude-code running
Runner spins up
Pod fires in your cluster, repo checked out warm, your Claude Max sub authorized into the agent CLI. Tokens stay on your side.
MR opens
Force-pushed branch, MR opened with a diff, CI status tracked back. You review, you merge. Same shape as Renovate.
What vemlor actually does
Lint errors, flaky tests, type fixes, codemod migrations. anything you can
describe in a @vemlor comment. Bring your Claude Code or Codex
subscription, we run the platform.
Agent activity across connected GitLab and GitHub workspaces
| # | Date | Status | Project | Cost |
|---|---|---|---|---|
| 1247 | 2026-06-05 | Merged | KMP gitlab.kompass.eu | 0.42 min |
| 1248 | 2026-06-05 | Running | KMP gitlab.kompass.eu | 1.20 min |
| 1246 | 2026-06-04 | Merged | NWN gitlab.northwind.eu | 0.18 min |
| 1245 | 2026-06-04 | Failed | STR gitlab.stratos.eu | 0.05 min |
Self-host on your cluster
Single Helm chart. Connect GitLab or GitHub via OAuth. Runners are ephemeral pods in your own Kubernetes cluster. Your tokens, your data.
# 1. install
$ helm install vemlor vemlor/vemlor
# 2. connect
$ vemlor connect gitlab https://gitlab.kompass.eu
✓ OAuth app linked
✓ Webhooks installed on 47 projects@vemlor in comments
Comment @vemlor fix the flake in auth_spec.rb on an MR. Webhook
fires, runner spins up in your cluster, agent edits, force-push, you review.
Force-pushed fix/auth-spec-flake. 3 retries stabilized. CI green.
Your sub. Your tokens. Our platform.
We don't resell LLMs. Your Claude Max or Codex subscription stays yours. We charge for the runners, the auth, the audit log, the control plane around it.
on tokens. You pay Anthropic or OpenAI direct, billed against your existing sub.
Claude Max sub powers the whole team's agent runs.
of fixes: lint, types, flakes, deps, codemods, anything in a @vemlor comment.
Start free. Stay flat.
Flat per-workspace pricing. No metered tokens, no surprise bills. Your Claude Max or Codex sub pays for the LLM.
- Everything in Solo
- 10 repos
- 1,000 autofixes / month
- Shared workspace · RBAC
- Audit log
- Email + Slack support
- Everything in Team
- 50 repos
- 5,000 autofixes / month
- Priority webhooks
- Autofix policy controls
Free for hobby projects · 1 repo, 30 autofixes/mo · Enterprise from $30k/yr · SSO, VPC, SLA. book a demo
Renovate's missing agent layer.
Renovate proved one shape: bot opens an MR, you review, you merge. Mend monetized Renovate Enterprise at the top of the market and ignored the mid-market self-hosted GitLab teams. That's where we live.
- renovate-style dep bumps
- flake repair
- codemod migrations
- lint fixes
- type tightening
What we do.
Anything you can describe in a @vemlor comment. Renovate-style
dep updates, flaky-test repair, codemod migrations, lint fixes,
type tightening. opened as an MR, on the platform you self-host.
What we don't do.
We don't resell tokens. We don't sell an IDE. We don't host your code. We don't compete on dep updates. Renovate already won that. We're the agent layer Renovate never built, behind your firewall, on your runners.
Renovate did dependency updates. vemlor does everything else.
Drop it onto your self-hosted GitLab. Bring your Claude Max or Codex sub. First MR opens in your inbox today.
@vemlor fix the flake in
spec/auth_spec.rb