GuidesSwitch to Work on Repeat

Migrate from Lindy

Move recurring Lindy agents into routines with explicit permissions — concept mapping, validation, cutover, and rollback.

This guide moves recurring Lindy agents — weekly research, reporting, and review work — into Work on Repeat routines. Conversational or event-reactive agents are out of scope; see Limitations.

Prerequisites

  • A Work on Repeat account and an available integration (or your own MCP server) for each tool the agent uses.
  • Access to the Lindy workspace, to read each agent's prompt, triggers, and action history.

Concept mapping

LindyWork on Repeat
Agent ("Lindy")Routine
Recurring triggerFive-field cron schedule with an IANA timezone
Skills / actionsConnected tools with per-tool permissions
Agent promptWritten instructions, versioned with the routine
Activity feedRun traces (every model step and tool call)

The key difference is permission posture: a routine's tools are individually enabled, and automatic execution is a separate grant per tool, so an unattended run can never use more than you explicitly allowed.

Export and back up

Copy each agent's prompt and note its triggers and connected accounts. Keep the agent enabled while you validate the routine against its recent activity.

Connections and permissions

Connect each provider, enable only the tools the responsibility needs, and keep mutations approval-gated. If Lindy held broad OAuth grants, reconnect with the narrowest scope the provider offers.

Routine setup and schedule translation

Recreate the agent's prompt as routine instructions — state the outcome, evidence expectations, and what the routine must never do. Translate the recurring trigger to cron (0 9 * * 1 for Monday 9am) with an explicit timezone.

Validate, cut over, roll back

Run manually, compare against the agent's recent outputs for a full cycle, then pause the Lindy agent rather than deleting it. Rollback is unpausing it.

Limitations

  • Prompts move by copy, not import; expect one editing pass to make instructions explicit about evidence and forbidden actions.
  • Chat-style, on-demand agents have no direct equivalent — routines run on a schedule or through the API.
  • Several managed integrations are connection-only previews with tools disabled until schema review.

The Lindy alternatives guide covers when Lindy remains the right tool.

On this page

Edit this page on GitHub