Skip to content

How Reminix works

Every run of a script gets its own fresh machine, which Reminix destroys when the run ends. A run sees only its version’s files and its inputs. It can’t see other runs, other scripts, your workspace’s data or Reminix itself. Reminix installs a version’s dependencies once, when you upload it, so a run never installs anything.

A run has up to 30 seconds. On Pro, a script can ask for up to 10 minutes, and on Business for up to 15. Such a long run returns right away and shows its progress as it goes. You can cancel it while it runs.

A script has no network access unless its version lists the hosts it may call. Even then, it can reach only those hosts, and only over HTTPS. A request to any other host gets an HTTP 403 right away, saying the host isn’t allowed. Reminix writes every request to the run’s log. The log holds up to 1,000 lines in total, including what the script writes there.

You save an API key once. Reminix encrypts it with a key for your workspace and never shows it again. Each secret lists the hosts Reminix may send it to. When a script’s request goes to one of those hosts, Reminix adds the secret on the way out, as a header such as Authorization: Bearer …. The code doesn’t hold the value, so it can’t log the secret or send it anywhere else. The one exception is a secret you let a script read as an environment variable. See Secrets.

Uploading code makes a draft version. Nothing changes for your team until someone publishes a version, which makes it the one runs use. The script’s owner or a workspace owner or admin can publish. You can test-run a draft before anyone publishes it, and rolling back is publishing an earlier version.

An AI agent’s publish through MCP always waits for a person to approve it. On Business, a workspace can also require a second person to approve every publish.

  • Who sees and runs a script: everyone in the workspace, or only chosen teams. Other members don’t see the script at all, whether in the app, through the API or as an agent’s tool.
  • Which runs wait for approval: none, the ones AI agents start, or every run (Business). A workspace owner or admin approves or denies each one. If nobody decides within a day, Reminix cancels the run.
  • Who manages secrets: workspace owners and admins. Members see a secret’s name and hosts, never its value. Agents connected through MCP can’t save or change secrets.

These are settings on the script, never part of its code, so publishing new code can’t loosen them. See Governance.

For each run, Reminix records its inputs, the version it used, who started it and where it came from: the app, the API, the CLI, an agent, a schedule or a trigger. It also records how the run ended, its output or error, and its log. Reminix keeps the run’s record. It keeps the run’s log and saved files for 7 days on Free, 30 on Pro and 90 on Business.

The workspace’s audit log records every change to access, policies, secrets and publishing.

If something interrupts a run, Reminix marks it failed. Reminix never retries a run on its own, because a retry could repeat what the script already did, such as sending an email twice.

Retrying the request that starts a run is safe. The SDKs, the CLI and the MCP server send an idempotency key and retry with the same key, so one call starts one run.