FAQ
Everything you need to know before you start. Still stuck? Email contact.automaintainer@protonmail.com.
The basics
What is automaintainer?
A service that keeps your repositories maintained automatically. You describe a focus in plain English (e.g. "keep tests passing and bump outdated dependencies"), and a team of AI agents plans the work, implements it, and opens a pull request — on a schedule you set.
How does it actually make changes?
For each run, agents work in an isolated checkout of your repo on a worker. They commit their work and open a pull request. Nothing reaches your default branch unless a PR is merged — and you control whether that's manual or automatic (see the trust ladder below).
What's a "focus"?
A short, plain-English objective for a repo. The best focuses are scoped and verifiable — "fix failing tests", "raise coverage toward 80%", "bump dependencies safely", "harden error handling", "keep the README in sync". Broad, open-ended asks ("build feature X") work less reliably — automaintainer is built for ongoing maintenance, not one-shot feature delivery.
How do I get started?
Sign up, add a repo (URL + focus), install a worker on a machine you control and plug in your AI provider key, then set a schedule. The dashboard gives you the exact install command. Your first PR shows up on the next run.
Safety & control
Will it break my code?
Changes only ever land as pull requests — never direct commits to your default branch. Each run is isolated, and you decide how much autonomy to grant via the trust ladder. On Free and Pro a human merges; even on Max, every change is a reviewable, revertable PR with full diff and CI.
What's the "trust ladder"?
Three levels of autonomy: Free — you review and merge. Pro — a separate AM review team reads the diff and checks CI, then you merge. Max — the review team also merges approved PRs for you, fully hands-off. You can change this per repository.
Is my code or data used to train models?
No. Your repositories are processed only to perform the maintenance runs you configure. (Note: token usage runs through your own AI provider under your account — see BYOK below — so your provider's data policy also applies.)
Where does my code actually run?
On your worker — a machine you control. Your code is cloned, edited, and committed there; the panel only orchestrates (scheduling, the agent workflow, surfacing logs and PRs). Your source never lives on our servers.
Access & security
What access does it need to my code?
A GitHub token you provide, used to clone, push branches, and open pull requests on the repositories you authorize — nothing else. Scope it to only the repos you want maintained. See our Security page.
Do private repositories work?
Yes. Access is entirely via the GitHub token you supply, so private repos work exactly like public ones.
How are secrets handled?
Secret values are never written to logs, PR bodies, or agent messages. Your AI provider keys stay on your worker (BYOK — see below); automaintainer never receives or stores them. Self-hosted deployments can source secrets from a dedicated encrypted secrets manager. See Security.
Pricing, plans & BYOK
Does my plan include AI / LLM usage costs?
No — automaintainer is bring-your-own-keys (BYOK). Your worker runs the AI coding harness (e.g. Claude Code, opencode) using your own provider API keys, configured on your machine. You pay your AI provider directly for token usage, and pay us only for orchestration. That's why plans are €19–€49/mo instead of the $200+/mo of bundled AI-engineer tools — no inference markup, and you keep full control of model choice and spend.
What do the plans include?
Orchestration: scheduling, the architect/dev/QA agent workflow, reviews, run history, and the dashboard — across a set number of repositories, workers, and runs per day. They do not include AI/LLM token costs (BYOK). See the pricing section on the home page.
Which AI models can I use?
Any model your harness supports — Claude, GPT, Gemini, open models via opencode, and more. You can even set a different model per agent role (architect / dev / QA) to balance cost and quality. Because it's BYOK, model choice and spend are yours.
What's the free trial?
Paid plans include a 48-hour free trial — no charge if you cancel before it ends. Free stays free forever, with no card required.
How do I cancel? Are there refunds?
Cancel anytime from the billing portal; access continues until the end of the current billing period. Fees already paid are non-refundable except where required by law. See Terms.
Workers & scheduling
What is a worker, and do I have to run one?
A worker is the machine that executes your runs (the am-cloud agent). You run it on your own infrastructure, which is what keeps your code and keys under your control. Every plan includes at least one worker; higher plans allow more for parallelism.
Does my machine need to stay on?
The worker must be online during a repo's scheduled window for that run to execute. Run it on an always-on box (a small VM or server) for reliable scheduled maintenance. If no worker is online when a run is due, that run is skipped/marked failed and retried on the next cycle.
Can I control when runs happen?
Yes — each repo has a schedule (days + hours, interpreted in Europe/Paris) and a minimum cooldown between runs.
What happens if a run fails?
It's recorded with its logs in the dashboard and simply retried on the next scheduled cycle. A failed run never touches your default branch.
Languages, platforms & scope
What languages and stacks are supported?
It's language-agnostic — the agents work in your repo the way a developer would (running your test command, your linter, etc.), so it adapts to whatever stack your project uses.
Which git platforms work?
GitHub today (clone, branches, pull requests, issue context). GitLab and Bitbucket aren't supported yet.
How are GitHub issues used?
When enabled, open issues are given to the agents as context so they can address them; reference Fixes #N in a commit and the issue auto-closes when the PR merges.
Will this work on a large codebase?
Yes — because each run is a scoped, verifiable objective, not a full-repo scan. The agent reads only the files relevant to the task, makes the change, and your verify command confirms nothing broke. For broader goals (e.g. "raise coverage to 70%"), AM stacks multiple focused runs over time — each one picking up where the last stopped via persistent repo memory. We shrink the objective, not the codebase.
Enterprise & support
Can I self-host the whole thing?
Yes — self-hosted deployment, SSO, audit logs, custom models/CLIs and an SLA are available on Enterprise. Contact us to discuss.
How do I get support?
Email contact.automaintainer@protonmail.com. Max and Enterprise plans get priority support.