What makes Crons different from a chat assistant
A chat assistant waits for you to ask. An agent on Crons does the work at the time you set, while you are in a meeting or asleep, and leaves the result where your team already works. That changes what you can hand over. Instead of asking an assistant every Monday to check competitor pricing, you describe the job once and read the Slack message when it arrives.
Agents on Crons are built from explicit steps, not a single long prompt. That makes them predictable. You can see which step fetched what, which step made a judgment, and which step sent a message. When something goes wrong, the run log points to the exact step.
What an agent can do inside a run
- Read the web with web fetch steps or a headless browser for pages that need clicks and scrolling.
- Call APIs with stored credentials, so the agent can query your own systems.
- Think with AI reasoning steps that classify, extract, compare and draft.
- Remember results between runs, so it reports what is new or what changed.
- Ask a person for approval before emailing customers, writing to a CRM or posting publicly.
- Deliver to Slack, Gmail, HubSpot, Google Sheets, Notion or any webhook.
Payments are never executed by an agent, on any plan. More on safeguards on the security page and the memory and approvals feature page.
Agents by department
Every department has recurring checks that eat a few hours a week. Here are agents teams commonly build first, each available as a template you can adjust.
| Department | Agent | Typical schedule | Delivers to |
|---|---|---|---|
| Sales | New lead finder with enrichment | Every weekday at 7 AM | HubSpot after approval |
| Marketing | SEO audit of key pages | Every Monday | Notion |
| Competitive intelligence | Competitor pricing and feature watch | Every Monday at 8 AM | Slack |
| Recruiting | Candidate sourcing shortlist | Every Tuesday and Thursday | Google Sheets |
| Finance | Invoice reconciliation | 1st of every month | Gmail draft for approval |
| Customer success | Customer health digest | Every Friday at 3 PM | Slack and Notion |
A good rule of thumb is to start with an agent that only reads and reports. Once the team trusts its summaries for a few weeks, add steps that write to tools, with an approval gate in front of them. That way the agent earns more responsibility the same way a new colleague would, and every decision it makes stays visible in the run log.
Explore these by team on the use cases page.
How to build an agent in Crons
- 1Describe the job in one sentence, including when it should run and where results should go.
- 2Review the parsed schedule, timezone and next three runs.
- 3Check the proposed steps and edit the instructions of any AI step.
- 4Connect the tools the agent reads from and delivers to.
- 5Place approval gates in front of any step that leaves your company.
- 6Run a dry run, read the sample result and save.
The whole process happens in the browser. There is nothing to host and no scheduler to maintain. For a longer walk through, read how to build AI agents on a schedule.
Choosing an AI agent builder
Many agent builders focus on conversation or on one off tasks. For recurring business work, look for five things: a real scheduler with timezones, memory across runs, retries for flaky sources, human approval for risky actions, and a run log you can read weeks later. Crons includes all five on every plan. Pro adds inbound webhooks, the REST API, an MCP server and code steps, so developers can call your agents from other systems. Teams adds SSO, an audit log, an IP allowlist for webhooks and a DPA.
Usage and limits
Each plan includes a monthly number of runs, AI calls and browser minutes, plus a daily cap on AI calls. Starter includes 3,000 AI calls a month, Pro 10,000 and Teams 35,000. Counters reset each billing period and deleting an agent does not return usage. You get an alert at 80 percent and the agent stops before exceeding the limit, so there are no surprise charges. See all numbers on the pricing page.
Frequently asked questions
What is an AI agent platform?
It is software for building, running and supervising AI agents that perform tasks with several steps, use tools and produce results without someone prompting each step. Crons specializes in agents that run on a schedule.
Can agents on Crons browse websites that need clicks?
Yes. Browser action steps run a headless browser that can click, scroll and read pages. Time spent in the browser counts toward your monthly browser minutes.
Will an agent ever take an action I did not approve?
Steps behind an approval gate wait until someone approves them. You decide which steps need approval, and payments are never executed by any agent.
Can developers call Crons agents from our own systems?
Yes, on Pro and Teams. You can trigger agents with inbound webhooks or the REST API, and connect through the MCP server.
How do I know what an agent did on a past run?
Open the run log. It records every step with its input, output, retries and timing, kept for 30 days on Starter, 90 days on Pro and one year on Teams.
Can I cancel my plan at any time?
Yes. Cancel in the billing portal and your agents keep running until the end of the paid period. Payments are final as described in the refund policy.