Automation FAQ

How do I build an automation?

You can build automations in three ways: using the Chat Builder (describe what you want in plain language and Chase Agents generates the steps), using the Visual Editor (add and configure steps manually using the drag-and-drop interface), or using the MCP execute_steps and create_automation tools from an external AI client like Claude Desktop.

What step types are available?

Chase Agents supports 14 step types: call_api_endpoint (call a REST API), run_sandboxed_code (run Python), use_mcp_tool (call a tool on an MCP connection), if (branch on a condition), switch (multi-way branch), loop (iterate over a list), merge (combine parallel branches), wait (pause execution), generate_pdf, generate_word, combine (combine step results), trigger_automation (run another automation), seek_human_approval_for_next_step (pause for human decision), and accept_external_input (wait for external data).

How do I pass data between steps?

Each step can access the output of the immediately preceding step via the dataPrev variable, and the outputs of all previous named steps by referencing their step name. In run_sandboxed_code steps, the data variable contains all available step outputs as a Python dictionary. The visual editor shows you the available data references when configuring each step.

Can I run Python code in an automation?

Yes. The run_sandboxed_code step executes Python in a secure sandbox. The sandbox includes common data processing libraries, the ai.generateText and ai.generateObject helper functions for LLM calls, rankByEmbeddings for semantic scoring, and document generation helpers including createExcel, createPDF, and createWordDoc. You cannot make outbound network requests from the sandbox; use a use_mcp_tool or call_api_endpoint step for external calls.

How do I trigger an automation on a schedule?

Set the automation's trigger type to Cron and provide a cron expression for the schedule you want. Chase Agents supports standard 5-field cron syntax. You can also use an Interval trigger for simpler recurring schedules (every N minutes or hours). The editor shows the next scheduled run time after you save.

How do I trigger an automation from an external system?

Set the automation's trigger type to Webhook. Chase Agents generates a unique webhook URL for the automation. Configure your external system to send a POST request to that URL when you want the automation to run. The request body is available to your automation steps as the trigger data.

How many automations can I run at once?

Each individual automation is rate-limited to 5 executions per 5-minute window. Multiple automations run independently and their rate limits do not affect each other. The number of automations that can be active simultaneously depends on your plan. See the Pricing page for current limits.

What happens if an automation fails?

The failure is logged in Workflow History with step-level detail. If the automation's error handling mode is set to ai, the platform attempts to automatically repair the failing step. If repair succeeds, the run continues. If not, the run is marked failed. You can view the error, fix the step in the visual editor, and re-run the automation manually.

Can one automation trigger another?

Yes. The trigger_automation step type lets you invoke another automation from within a running automation, optionally passing data to it. This is useful for modular workflow design where a parent automation orchestrates several specialized child automations.