Monitors

A monitor is a check Manifold repeats on a schedule. Each check makes the same tool calls your agent could make once, keeps what it saw, and compares it with the last check, so it can report only what changed: a new Reddit thread, a price on a competitor's page, an ad that started, a ranking that moved, a brand that appeared in an AI answer. You can make a monitor in the dashboard, or ask your agent to make one. The agent reads what monitors find in any connected client.

For what each kind of monitor is for, with prices, see the monitor pages.

Monitor types

Agents can also make a Reviews monitor, for new reviews of an app, a Google Maps place or a Google Shopping product. The dashboard does not offer it.

How checks run

• Hourly, daily or weekly. Each type starts at the default above, and you can change it. Checks repeat from the first one, with no day or time to pick.
• The first check is a starting point. It runs within five minutes, records what is there and reports nothing, so a week of old posts never arrives as news. Changes are reported from the second check on.
• Checks read live data. They skip the cache, because a check that read an hours-old answer would find nothing new, and they pay the full price of each call.
• A failed check is not retried, since a retry would charge again. It is recorded, and the next check runs on schedule.
• Each workspace can keep 25 monitors.

What it costs

Creating, reading and changing monitors is free. Each check charges for the calls it makes, at the same credit prices as your agent's calls. The form, and the agent's preview, show the monthly estimate before the monitor starts. Some examples, worked out the same way:

A monitor has no credit limit of its own. Your workspace's spend cap applies to monitors as to everything else: at the cap, checks are skipped until it resets. With no credits left, a monitor pauses until you add credits or a plan renews.

Alerts

A change that cannot wait is sent as an email alert: at most one a day per monitor, with the rest waiting in its feed. Alerts start on where waiting costs something, such as a Reddit thread, which is answered in its first day or two. AI answers and rankings move on noise, so their alerts start off, and when on, they wait for a move that holds. Each person in the workspace turns a monitor's alerts on or off for themselves.

Your agent and monitors

Ask in any connected client, such as “monitor Reddit for our brand every hour” or “what did my monitors find this week?”. The agent shows the monthly estimate with monitoring_create_monitor in preview first, and makes the monitor once you agree. It reads finds with monitoring_get_changes, the tracked number over time with monitoring_get_history, and pauses, changes or runs a check now with monitoring_update_monitor. These tools are free; every parameter and return field is on the monitoring tools page.

Give each monitor a description saying what it is for. An agent that reads its changes later uses it to decide what to do with them, so one is written from the fields if you leave it out.

What is kept

A change keeps its summary for as long as the monitor exists. Its quote, such as a post or page title, and its details, such as the author, likes and views, are cleared after 30 days. An ad's text is kept until 30 days after the ad was last seen, so the monitor can list the ads still running. Whole posts and AI answer text are never stored.