lindy-upgrade-migration

What this skill does

# Lindy Upgrade & Migration

## Overview

Lindy is a managed platform — agents run on Lindy's infrastructure. "Upgrades"
mean reconfiguring agents for new capabilities, migrating agents between workspaces,
or adapting to platform changes. Key concern: agents with webhooks, Lindymail,
and phone numbers require reconfiguration after migration.

## Prerequisites

- Admin access to source and target Lindy workspaces
- Inventory of all agents, triggers, and integrations
- Migration window scheduled for customer-facing agents

## Instructions

### Step 1: Inventory Current Agents

Document every agent before making changes:

| Agent Name | Trigger Type | Actions | Integrations | Webhook URL | Phone # |
|-----------|-------------|---------|--------------|-------------|---------|
| Support Bot | Email Received | Gmail Reply, Slack Notify | Gmail, Slack | N/A | N/A |
| Lead Router | Webhook | Sheets Update, Slack DM | Sheets, Slack | `https://public.lindy.ai/...` | N/A |
| Phone Screener | Call Received | Transfer, Agent Send | Phone | N/A | +1-555-0100 |

### Step 2: Export Agent Configurations

For each agent, document:

- **Prompt**: Copy full text from Settings > Prompt
- **Model**: Which AI model is selected
- **Skills/Actions**: List all action steps and their configurations
- **Trigger filters**: Copy filter conditions
- **Knowledge Base**: Note all sources (files, URLs, integrations)
- **Memories**: Export any persistent memories
- **Exit conditions**: Copy all condition text

### Step 3: Plan Migration Order

```
Phase 1: Internal-only agents (no customer impact)
  → Migrate, test, verify for 24-48 hours

Phase 2: Low-risk customer-facing agents (email triage, notifications)
  → Migrate during low-traffic window
  → Monitor for 24 hours

Phase 3: Critical agents (phone, live chat, lead routing)
  → Migrate with rollback plan ready
  → Keep old agent active in parallel for 48 hours
```

### Step 4: Migrate Agent to New Workspace

**Option A — Template sharing**:

1. In source workspace: Share agent as **Template**
2. In target workspace: Import template
3. Reconfigure integrations (OAuth tokens are NOT transferred)
4. Re-authorize all connected services

**Option B — Manual recreation**:

1. Create new agent in target workspace
2. Paste saved prompt
3. Recreate trigger with same configuration
4. Re-add all actions and configure fields
5. Upload knowledge base files
6. Re-create memories

### Step 5: Reconfigure Webhooks, Email, and Phone

These require special attention — they are NOT automatically transferred:

**Webhooks**:

- New agent gets a NEW webhook URL
- Update all calling systems with the new URL
- Generate new webhook secret
- Update all clients with new Bearer token

**Lindymail** (Lindy-assigned email addresses):

- New agent gets a new Lindymail address
- Update forwarding rules and any published email addresses

**Phone numbers**:

- Phone numbers may need to be re-provisioned ($10/month each)
- Update IVR systems and published phone numbers
- Test call quality and language settings

### Step 6: Parallel Run & Cutover

```
Day 1-2: Both old and new agents active
  → Route test traffic to new agent
  → Compare task completion rates and output quality

Day 3: Gradual cutover
  → Redirect 50% of traffic to new agent
  → Monitor error rates and credit consumption

Day 4: Full cutover
  → Route 100% to new agent
  → Keep old agent paused (not deleted) for 7 days

Day 11: Cleanup
  → Delete old agent after 7-day safety window
```

### Step 7: Verify Post-Migration

- [ ] All triggers firing correctly
- [ ] All actions completing successfully
- [ ] Knowledge base returning relevant results
- [ ] Webhook URLs updated in all calling systems
- [ ] Phone numbers tested (inbound and outbound)
- [ ] Credit consumption within expected range
- [ ] Team members have correct access in new workspace

## Common Migration Scenarios

### Scenario: Consolidating Multiple Agents

When you have too many single-purpose agents:

1. Identify agents with overlapping triggers
2. Merge prompts into sections (use `## Billing`, `## Technical`, etc.)
3. Add conditions to route based on content
4. Reduce total active agents → lower costs

### Scenario: Upgrading Agent Capabilities

When Lindy adds new features (new actions, new models):

1. Review [Lindy Changelog](https://www.lindy.ai/changelog)
2. Update model selection if better options available
3. Replace workaround actions with new native actions
4. Test thoroughly — model changes affect output quality

### Scenario: Environment Promotion (Dev to Prod)

1. Share dev agent as Template
2. Import in production workspace
3. Re-authorize all production integrations
4. Update webhook URLs to production endpoints
5. Verify trigger filters match production requirements

## Error Handling

| Issue | Cause | Solution |
|-------|-------|----------|
| Webhook URL changed | New agent gets new URL | Update all callers with new URL |
| Integration auth failed | OAuth not transferred | Re-authorize in new workspace |
| Knowledge base empty | Files not re-uploaded | Upload files to new agent's KB |
| Phone number unavailable | Not re-provisioned | Purchase new number in settings |
| Memory lost | Memories not exported | Manually re-create critical memories |

## Resources

- [Lindy Changelog](https://www.lindy.ai/changelog)
- [Lindy Templates](https://docs.lindy.ai/fundamentals/lindy-101/templates)
- [Lindy Documentation](https://docs.lindy.ai)

## Next Steps

Proceed to Pro tier skills for advanced CI integration, deployment, and performance.