Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Advanced migration strategies for moving to Lindy AI from other platforms. Use when migrating from Zapier, Make, n8n, custom automations, or consolidating fragmented agent systems. Trigger with phrases like "lindy migration", "migrate to lindy", "zapier to lindy", "switch to lindy", "consolidate automations".
.claude/skills/jeremylongshore-lindy-migration-deep-dive/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 44% | 0% |
Migrate existing automation workflows from Zapier, Make (Integromat), n8n, LangChain, or custom code to Lindy AI. Key insight: Lindy replaces rigid rule-based automations with AI agents that can reason, adapt, and handle ambiguity — so migration is a redesign opportunity, not a 1:1 translation.
| Source Platform | Lindy Equivalent | Key Difference | |----------------|-----------------|----------------| | Zapier Zap | Lindy Agent | AI reasoning replaces rigid if/then | | Make Scenario | Lindy Agent | No-code builder instead of module chains | | n8n Workflow | Lindy Agent | Managed infra, no self-hosting | | LangChain Agent | Lindy Agent Step | No-code, managed, no Python needed | | Custom code | HTTP Request + Run Code | Less code, AI fills gaps |
For each existing automation, document:
| Field | Example | |-------|---------| | Name | Support Email Triage | | Trigger | New email in support@co.com | | Steps | 1. Parse email 2. Classify 3. Route to channel | | Integrations | Gmail, Slack, Sheets | | Frequency | ~50 runs/day | | Complexity | Medium (3 steps, 1 condition) |
| Complexity | Criteria | Migration Approach | Time | |-----------|---------|-------------------|------| | Simple | 1-3 steps, no conditions | Build from scratch in Lindy | 30 min | | Medium | 4-8 steps, conditions | Natural language description to Agent Builder | 1-2 hours | | Complex | 9+ steps, multi-branch, loops | Redesign as multi-agent society | 1-2 days | | Custom code | Python/JS logic | Run Code action + HTTP Request | 2-4 hours |
From Zapier:
Zapier Pattern → Lindy Pattern
────────────────────────────────
Trigger (New Email) → Trigger (Email Received)
Filter Step → Trigger Filter (more efficient)
Formatter → AI Prompt field mode (AI does formatting)
Lookup → Knowledge Base search or HTTP Request
Multi-step Zap → Single agent with conditions
Paths → Conditions (natural language branching)From Make (Integromat):
Make Pattern → Lindy Pattern
────────────────────────────────
Scenario → Agent workflow
Module → Action step
Router → Conditions
Iterator → Loop
Aggregator → Run Code action (consolidation logic)
Error Handler → Agent prompt error instructionsFrom n8n:
n8n Pattern → Lindy Pattern
────────────────────────────────
Trigger Node → Trigger
Function Node → Run Code (Python/JS)
HTTP Request Node → HTTP Request action
IF Node → Condition
Merge Node → Agent step (AI merges intelligently)From LangChain/Custom Code:
LangChain Pattern → Lindy Pattern
────────────────────────────────
Agent → Agent Step with skills
Tool → Action or HTTP Request
Memory → Lindy Memory (persistent)
Chain → Workflow steps
Vector Store → Knowledge Base
Retrieval Chain → Knowledge Base + AI PromptPhase 1: Internal-Only Agents (Days 1-3)
Phase 2: Low-Risk Customer-Facing (Days 4-7)
Phase 3: Critical Workflows (Days 8-14)
Migration is a chance to improve, not just replicate:
| Old Pattern | Lindy Improvement | |------------|-------------------| | Rigid if/then classification | AI classifies naturally, handles edge cases | | Template-based email responses | AI generates contextual, personalized responses | | Multiple automations for variations | Single agent with conditions handles all | | Manual data transformation | Run Code action or AI handles transformation | | No error handling | Agent prompt includes fallback behavior |
bash# Post-migration validation checklist echo "=== Migration Validation ===" # 1. Task completion rate echo "Check: Agent Tasks tab - expect >95% success rate" # 2. Response quality echo "Check: Compare 10 agent outputs to old automation outputs" # 3. Trigger coverage echo "Check: All events triggering correctly (no missed events)" # 4. Performance echo "Check: Task duration within acceptable range" # 5. Cost echo "Check: Credit consumption within budget"
| Issue | Cause | Solution | |-------|-------|----------| | Output quality lower | AI prompt needs tuning | Add few-shot examples to agent prompt | | Missing edge cases | Source had specific rules | Add condition branches or prompt instructions | | Higher cost than expected | Overuse of large models | Right-size models per step | | Integration auth fails | OAuth not set up in Lindy | Authorize integrations before migration | | Data format mismatch | Different field names | Map fields in Run Code action |
This completes the Flagship tier. Review Standard and Pro skills for comprehensive Lindy mastery.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 47,359 | 38,203 | -19% | 1 | 1 | 0% | 6,810 | 6,922 | +2% | 0 | 0 | — |
case-02 | fail→fail | 41,748 | 36,748 | -12% | 1 | 1 | 0% | 5,893 | 6,967 | +18% | 0 | 0 | — |
case-03 | fail→pass | 33,852 | 34,689 | +2% | 1 | 1 | 0% | 4,782 | 6,834 | +43% | 0 | 0 | — |
case-04 | fail→pass | 18,908 | 15,864 | -16% | 1 | 1 | 0% | 2,290 | 4,174 | +82% | 0 | 0 | — |
case-05 | fail→pass | 25,848 | 24,063 | -7% | 1 | 1 | 0% | 3,145 | 4,930 | +57% | 0 | 0 | — |
case-06 | fail→fail | 19,383 | 22,949 | +18% | 1 | 1 | 0% | 2,643 | 4,137 | +57% | 0 | 0 | — |
case-07 | fail→pass | 14,674 | 12,135 | -17% | 1 | 1 | 0% | 1,994 | 2,865 | +44% | 0 | 0 | — |
case-08 | fail→pass | 15,805 | 18,466 | +17% | 1 | 1 | 0% | 2,084 | 3,425 | +64% | 0 | 0 | — |
case-09 | pass→pass | 16,693 | 10,835 | -35% | 1 | 1 | 0% | 1,785 | 3,136 | +76% | 0 | 0 | — |
case-10 | pass→pass | 22,299 | 12,967 | -42% | 1 | 1 | 0% | 2,184 | 3,404 | +56% | 0 | 0 | — |
case-11 | pass→pass | 22,128 | 18,977 | -14% | 1 | 1 | 0% | 2,678 | 4,059 | +52% | 0 | 0 | — |
case-12 | fail→pass | 18,111 | 8,395 | -54% | 1 | 1 | 0% | 2,131 | 2,259 | +6% | 0 | 0 | — |
case-13 | fail→pass | 11,516 | 2,725 | -76% | 1 | 1 | 0% | 1,842 | 2,041 | +11% | 0 | 0 | — |
case-14 | fail→pass | 12,893 | 16,771 | +30% | 1 | 1 | 0% | 1,318 | 3,213 | +144% | 0 | 0 | — |
case-15 | pass→pass | 17,320 | 11,604 | -33% | 1 | 1 | 0% | 1,774 | 2,656 | +50% | 0 | 0 | — |
case-16 | fail→pass | 21,501 | 7,213 | -66% | 1 | 1 | 0% | 2,560 | 2,048 | -20% | 0 | 0 | — |
case-17 | pass→pass | 17,670 | 10,760 | -39% | 1 | 1 | 0% | 2,021 | 3,374 | +67% | 0 | 0 | — |
case-18 | fail→pass | 18,208 | 12,340 | -32% | 1 | 1 | 0% | 1,998 | 3,601 | +80% | 0 | 0 | — |
case-19 | fail→pass | 11,450 | 9,746 | -15% | 1 | 1 | 0% | 1,752 | 2,480 | +42% | 0 | 0 | — |
case-20 | pass→pass | 17,416 | 25,180 | +45% | 1 | 1 | 0% | 2,274 | 5,256 | +131% | 0 | 0 | — |
case-21 | pass→pass | 26,128 | 15,247 | -42% | 1 | 1 | 0% | 2,893 | 3,573 | +24% | 0 | 0 | — |
case-22 | pass→pass | 19,892 | 19,224 | -3% | 1 | 1 | 0% | 2,391 | 3,823 | +60% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +55 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.