Verdict, October 23, 2026. OP to turn on monitoring alerts and Auto-fix for a published Replit app, then read the result in Agent chat before you republish. Mid to turn on Auto-republish. Replit marks that switch experimental, says it requires Auto-fix, and says some changes, including migrations and index operations, may be risky. No letter moves on the AI Coding Agents board. Replit is not a row on it. S stays locked. OP or Mid has not published an app with these switches on.
The October 22, 2026 launch post and the Monitoring Agent docs, read October 23, 2026. Behavior described there is a vendor claim.
Three switches, not one agent
NEW Replit published Monitoring Agent on October 22, 2026. The post says it can investigate an alert, identify the cause, validate a fix, and deploy that fix. The same post says you decide whether fixes are published automatically. Source: launch post (opens in a new tab).
The docs split that into three settings. Alerts do not let Agent change your code. Auto-fix attempts a fix and does not authorize a live publish. Auto-republish, which the launch post calls auto-publish, can publish a validated fix. We did not open the Publishing pane, so those two labels are matched from the descriptions, not from a screenshot. Sources: Monitoring Agent (opens in a new tab), launch post (opens in a new tab).
| Setting | What the docs say it does | Live app |
|---|---|---|
| Monitoring alerts | Records production issues. On an existing app, the switch saves without republishing | Unchanged |
| Auto-fix | Investigates in the background, attempts a fix, and validates it | Unchanged until you republish |
| Auto-republish | Publishes that validated fix. Experimental. Requires Auto-fix | Changes |
Source: Monitoring Agent (opens in a new tab).
NERF An alert does not guarantee a fix. The docs' example is an external service that needs action outside the app. Source: Monitoring Agent (opens in a new tab).
Automatic checks watch three signals: application error logs, HTTP errors, and request latency. Those checks do not cover every metric Agent can investigate. Alerts Replit cannot evaluate show Cannot measure. Active means the issue still affects the app. Resolved means it stopped within the last 30 days. Source: Monitoring (opens in a new tab).
The fix: alerts and Auto-fix, then you publish
BUFF OP for a published app you can review. Each app has its own settings.
- Open the Publishing tool. For an app that is already live, open its publish settings. For a first publish, set monitoring there. Source: Monitoring Agent (opens in a new tab).
- Set a usage limit before Auto-fix. The docs say Monitoring Agent runs may incur Agent usage charges. On a Core account: Settings, then Usage, then Manage limits under Account usage. For an organization, open Usage and use Manage limits there. A usage limit blocks usage-based services past the cap until the next cycle or until you raise it. Source: Monitoring Agent (opens in a new tab), Managing your spend (opens in a new tab).
- Switch on Monitoring alerts. On an existing app that saves immediately, with no republish. On a first publish, Enable Active monitoring under Suggested add-ons also enables uptime checks where available. Uptime checks are separate, require a paid plan, and are not available for Scheduled Deployments. The docs do not say monitoring alerts themselves require a paid plan. Sources: Monitoring Agent (opens in a new tab), Monitoring (opens in a new tab).
- Switch on Auto-fix. An existing app saves immediately. A new app needs Publish before the setting applies and monitoring starts. Source: Monitoring Agent (opens in a new tab).
- Leave Auto-republish off. Review the result in Agent chat, then select Publish again. The app updates at the same URL. Visitors get the new version the next time they load it. Sources: Monitoring Agent (opens in a new tab), After you publish (opens in a new tab).
QoL You can ask Agent about production without Auto-fix and without waiting for an alert. The docs' examples cover last-hour 500s, today's slowest endpoints with p50 and p99, CPU and memory since the last publish, and logs for a revision that failed to start. Name the time, time zone, and endpoint when you know them. On-demand metrics go past the three alert signals, and which ones exist depends on the deployment type and cloud provider. Source: Monitoring Agent (opens in a new tab).
NERF Turning off alert email does not turn off monitoring or Auto-fix. Under Settings, then Notifications, Active monitoring alerts can be Critical only, High and above, Medium and above, all levels, or no emails. Production alert emails have to be enabled or the severity selector stays disabled. Authorized admins set account recipients separately: Off, Admins, or Custom, with their own severity threshold. The launch post says that with alerting and autofix on, Agent emails a summary of what happened. The docs page read here does not describe that summary. Sources: Monitoring Agent (opens in a new tab), launch post (opens in a new tab).
Why Auto-republish stays Mid
NERF Mid, and leave it off, until you have read Auto-fix diffs on that app. The docs call Auto-republish experimental, require Auto-fix, and say Auto-fix alone does not authorize publishing to the live app. The same note says some changes, including migrations and index operations, may be risky. It does not say Agent refuses those changes. Source: Monitoring Agent (opens in a new tab).
A manual republish runs five stages: Provision, Security checks, Build, Bundle, and Promote. Security checks scan dependencies and code, and can block a publish when critical vulnerabilities are found, depending on your security settings. The Monitoring Agent page does not say an automatic republish uses that pipeline or skips it. Source: After you publish (opens in a new tab).
If a publish succeeds and the live app returns an internal server error, the deployment help page says to open Publishing, then History, find the last working version, and redeploy it. That is the general recovery path, not a Monitoring Agent undo. Source: Deployment and publishing (opens in a new tab).
PRICE Paid Agent work is effort-based, and a checkpoint is completed work on a request. The billing page says Replit asks for confirmation before a paid action starts. The Monitoring Agent page does not say a background Auto-fix asks first, which mode it uses, or what one alert costs. Usage can take up to 30 minutes to show on the dashboard. Core lists one active background task, Pro lists ten, and Starter lists zero. Auto-fix "responds to alerts in the background." The docs do not say that run counts against this cap. Sources: AI billing (opens in a new tab), Monitoring Agent (opens in a new tab).
The pricing page, read October 23, 2026, shows Core at $18 a month billed annually, with $20 struck through on that card, and the line "$20 towards more powerful models." Pro shows $90 a month billed annually, with $100 struck through, and "$100 towards most powerful models." Prices are subject to tax. Plan docs point at that page for the monthly figure, so this article does not state one. Starter includes one published app, and that link goes down after 30 days. Sources: pricing (opens in a new tab), Replit Core (opens in a new tab), Starter plan (opens in a new tab).
The board stays put
No letter moves. Codex, Claude Code, Antigravity, and Cursor stay in A on the October 15 AI Coding Agents list. S stays locked. Monitoring Agent is a switch on a hosted app, not a new harness rank. The Replit row is a directory listing checked October 23, 2026, not a tier.
What to do today
- Set a usage limit, then turn on Monitoring alerts and Auto-fix for each published app you can review.
- Leave Auto-republish off. Read the chat, then Publish yourself. The URL stays the same.
- Do not treat a silenced inbox as off. Email settings do not stop Auto-fix.
- If a live release returns a 500, redeploy the last working version from Publishing, then History.
- Ask Agent about a time window before you wait for an alert. Name the time zone and the endpoint.
The weekly patch is on beehiiv: subscribe free (opens in a new tab).
The OP or Mid Team
Sources
Checked October 23, 2026.
- Replit Monitoring Agent, October 22, 2026 (opens in a new tab)
- Monitoring Agent (opens in a new tab)
- Monitoring your app (opens in a new tab)
- After you publish (opens in a new tab)
- Deployment and publishing (opens in a new tab)
- Replit AI billing (opens in a new tab)
- Managing your spend (opens in a new tab)
- Pricing (opens in a new tab)
- Replit Core (opens in a new tab)
- Starter plan (opens in a new tab)