Migrate from Lindy to Manus.
2 documentation-derived translation patterns — what carries over and what to watch for. Cited to the Feature Parity Map; the audit tells you whether the move is worth it.
Both tools take a multi-step goal in plain English and execute the whole workflow on the user's behalf — Lindy by assembling a saved 'Lindy' (pick a trigger, connect apps, give natural-language instructions), Manus by handing one prompt to a general agent that plans and runs the task inside its own cloud computer (browser, file system, code execution). To cut Lindy and consolidate onto Manus a team already uses, stop rebuilding agents in Lindy's visual builder and instead describe the same outcome to Manus as a task: state the goal, the inputs, and the apps to touch, and let Manus plan and execute, watching the step-by-step view and taking over or stopping if needed. For Lindy automations that ran on a recurring trigger (e.g. 'every morning' or 'on new email'), recreate the cadence with a Manus scheduled/recurring task rather than an always-listening trigger, and rewire each app Lindy connected via Manus Connectors (node:manus/connectors) — confirm each one exists before retiring the Lindy version. Keep Manus as the single agent surface; cut Lindy.
- Warning: Model mismatch: Lindy is an event-driven 'always-on' assistant that wakes on a trigger and reacts (new email, inbound message); Manus is task/session-oriented and runs a goal to completion. Reactive, always-listening Lindy flows (auto-reply to every inbound email the moment it lands) do not map cleanly to a one-shot Manus task — re-scope them as scheduled polling runs or accept they may not have a 1:1 equivalent before cancelling Lindy.
- Warning: Export first: Lindy holds your agents, prompts, templates, and per-app connections inside its builder with no portable agent-export format — copy out each agent's instructions, trigger config, and connected accounts before you cancel, because there is no import path into Manus.
- Warning: Re-authorize every integration: Manus reaches apps through its own Connectors/MCP layer (OAuth, sometimes a personal API token), so each Gmail/Calendar/CRM/Notion link Lindy had must be reconnected in Manus — and a few niche apps Lindy supported may have no Manus connector yet, needing a custom MCP server or a browser-driven fallback.
- Warning: Billing model differs: Lindy charges on task/credit tiers per assistant; Manus meters every run in credits that reset monthly and do not roll over, with heavier multi-step tasks costing more — model the team's real task volume against Manus credits so the consolidation does not silently cost more than the Lindy plan it replaces.
Both tools fan one big job out across many agents working at the same time: Lindy 'agent swarms' let a single agent duplicate itself to handle each row of a list/spreadsheet in parallel (e.g. spawn hundreds of agents to research a lead and draft a personalized email each), and Manus 'Wide Research' splits a complex task across parallel sub-agents — each in its own VM with its own browser and internet — and a main agent collects the results. To retire a Lindy swarm and run the same fan-out in Manus a team already has, hand Manus the whole batch as one task and describe the per-item work plus the desired combined output; when Manus judges the task is decomposable it auto-launches Wide Research and parallelizes it for you. Keep Manus as the parallel-execution surface; cut Lindy.
- Warning: No manual trigger and a hard width cap: Wide Research only kicks in automatically when Manus decides the task is parallelizable — you cannot force it — and it runs up to ~20 sub-tasks at once (framed as 20 agents on one task), with each sub-task capped at 50 credits. Lindy swarms that spawned hundreds of parallel agents over a large spreadsheet will not map 1:1; expect Manus to batch/serialize beyond its width, so re-scope very large fan-outs (or chunk the input) before assuming parity, and verify on a small batch first.
- Warning: Paid-only on both ends, metered on Manus: Wide Research is available to paid Manus users only and every sub-agent consumes credits that reset monthly and do not roll over — a large parallel run can burn a real chunk of the monthly allocation, so size the team's credit plan against the batch volume the Lindy swarm used to handle before cancelling Lindy.
- Warning: Isolated sub-agents, not a coordinated workflow: Manus sub-agents run independently and do not talk to each other — the main agent only distributes and reassembles. Lindy swarm patterns that depended on agents handing off or sharing state mid-run (one qualifies, another follows up, a third updates the CRM) are a sequential multi-agent workflow, not Wide Research; rebuild those as an ordinary multi-step Manus task (node:manus/autonomous-agent), not as a parallel fan-out.