Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Set up local development workflow for testing Lindy AI agent integrations. Use when building webhook receivers, testing agent callbacks, or iterating on Lindy-connected applications locally. Trigger with phrases like "lindy local dev", "lindy development", "test lindy locally", "lindy webhook local".
.claude/skills/jeremylongshore-lindy-local-dev-loop/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 107% | 0% |
Lindy agents run on Lindy's managed infrastructure — you do not run agents locally. Local development focuses on building and testing the webhook receivers, callback handlers, and application code that Lindy agents interact with. Use ngrok or similar tunnels to expose local endpoints for Lindy webhook triggers.
lindy-install-auth setuptypescript// server.ts — Express webhook receiver for Lindy callbacks import express from 'express'; import dotenv from 'dotenv'; dotenv.config(); const app = express(); app.use(express.json()); const WEBHOOK_SECRET = process.env.LINDY_WEBHOOK_SECRET; // Verify Lindy webhook authenticity function verifyWebhook(req: express.Request): boolean { const auth = req.headers.authorization; return auth === `Bearer ${WEBHOOK_SECRET}`; } // Receive Lindy agent callbacks app.post('/lindy/callback', (req, res) => { if (!verifyWebhook(req)) { console.error('Unauthorized webhook attempt'); return res.status(401).json({ error: 'Unauthorized' }); } console.log('Lindy callback received:', JSON.stringify(req.body, null, 2)); // Process the agent's output const { taskId, result, status } = req.body; console.log(`Task ${taskId}: ${status}`); res.json({ received: true }); }); // Health check for Lindy to verify endpoint app.get('/health', (req, res) => res.json({ status: 'ok' })); app.listen(3000, () => console.log('Webhook receiver running on :3000'));
bash# Install and start ngrok npm install -g ngrok ngrok http 3000 # Output: https://abc123.ngrok.io -> http://localhost:3000 # Use this URL in Lindy webhook configuration
In the Lindy dashboard, add an HTTP Request action to your agent:
https://abc123.ngrok.io/lindy/callbackContent-Type: application/json Send the task result as JSON with fields: taskId, result, status
Or configure a webhook trigger pointing to your tunnel URL:
https://abc123.ngrok.io/lindy/webhooktypescript// test-trigger.ts — Fire a test webhook to your Lindy agent import fetch from 'node-fetch'; async function triggerAgent() { const WEBHOOK_URL = process.env.LINDY_WEBHOOK_URL!; const SECRET = process.env.LINDY_WEBHOOK_SECRET!; const response = await fetch(WEBHOOK_URL, { method: 'POST', headers: { 'Authorization': `Bearer ${SECRET}`, 'Content-Type': 'application/json', }, body: JSON.stringify({ action: 'test', data: { message: 'Hello from local dev', timestamp: new Date().toISOString() }, }), }); console.log(`Status: ${response.status}`); console.log(`Response: ${await response.text()}`); } triggerAgent();
json// package.json scripts { "scripts": { "dev": "tsx watch server.ts", "test:trigger": "tsx test-trigger.ts", "tunnel": "ngrok http 3000" } }
bash# Terminal 1: Start server with auto-reload npm run dev # Terminal 2: Start tunnel npm run tunnel # Terminal 3: Fire test triggers npm run test:trigger
bash# .env LINDY_API_KEY=lnd_live_xxxxxxxxxxxx LINDY_WEBHOOK_SECRET=whsec_xxxxxxxxxxxx LINDY_WEBHOOK_URL=https://public.lindy.ai/api/v1/webhooks/YOUR_ID NODE_ENV=development
[Edit local code] → [Auto-reload via tsx watch]
↓
[Fire test webhook] → [Lindy agent processes]
↓
[Agent calls back] → [ngrok tunnel → localhost:3000]
↓
[Review logs] → [Iterate]| Issue | Cause | Solution | |-------|-------|----------| | ngrok tunnel expires | Free tier limit (2hr) | Restart ngrok or use paid plan | | Lindy can't reach endpoint | Tunnel URL changed | Update webhook URL in Lindy dashboard | | Callback not received | Agent HTTP Request misconfigured | Check URL and headers in action config | | ECONNREFUSED | Local server not running | Start server before testing | | SSL error | ngrok not using HTTPS | Always use the https:// ngrok URL |
Proceed to lindy-sdk-patterns for integration patterns and best practices.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,506 | 16,347 | -27% | 1 | 1 | 0% | 3,595 | 3,826 | +6% | 0 | 0 | — |
case-02 | fail→pass | 25,045 | 26,717 | +7% | 1 | 1 | 0% | 4,056 | 5,079 | +25% | 0 | 0 | — |
case-03 | fail→pass | 29,242 | 27,784 | -5% | 1 | 1 | 0% | 4,275 | 4,896 | +15% | 0 | 0 | — |
case-04 | pass→pass | 22,267 | 15,742 | -29% | 1 | 1 | 0% | 3,369 | 3,559 | +6% | 0 | 0 | — |
case-05 | pass→pass | 15,918 | 14,455 | -9% | 1 | 1 | 0% | 1,893 | 2,971 | +57% | 0 | 0 | — |
case-06 | fail→pass | 17,703 | 8,010 | -55% | 1 | 1 | 0% | 2,167 | 1,914 | -12% | 0 | 0 | — |
case-07 | pass→pass | 15,267 | 11,692 | -23% | 1 | 1 | 0% | 1,862 | 2,300 | +24% | 0 | 0 | — |
case-08 | pass→pass | 13,409 | 10,215 | -24% | 1 | 1 | 0% | 1,573 | 2,341 | +49% | 0 | 0 | — |
case-09 | pass→pass | 13,956 | 7,662 | -45% | 1 | 1 | 0% | 1,571 | 2,448 | +56% | 0 | 0 | — |
case-10 | fail→fail | 15,852 | 11,744 | -26% | 1 | 1 | 0% | 2,035 | 3,176 | +56% | 0 | 0 | — |
case-11 | pass→pass | 13,384 | 18,167 | +36% | 1 | 1 | 0% | 2,214 | 3,135 | +42% | 0 | 0 | — |
case-12 | pass→pass | 26,340 | 11,072 | -58% | 1 | 1 | 0% | 2,529 | 3,379 | +34% | 0 | 0 | — |
case-13 | fail→fail | 20,133 | 11,096 | -45% | 1 | 1 | 0% | 2,672 | 3,195 | +20% | 0 | 0 | — |
case-14 | fail→pass | 14,729 | 9,239 | -37% | 1 | 1 | 0% | 2,102 | 2,917 | +39% | 0 | 0 | — |
case-15 | pass→pass | 14,251 | 3,760 | -74% | 1 | 1 | 0% | 1,866 | 1,879 | +1% | 0 | 0 | — |
case-21 | pass→pass | 10,826 | 4,243 | -61% | 1 | 1 | 0% | 1,173 | 1,965 | +68% | 0 | 0 | — |
case-16 | pass→pass | 9,180 | 9,676 | +5% | 1 | 1 | 0% | 1,176 | 2,236 | +90% | 0 | 0 | — |
case-17 | pass→pass | 17,343 | 10,410 | -40% | 1 | 1 | 0% | 1,823 | 2,208 | +21% | 0 | 0 | — |
case-18 | fail→fail | 17,357 | 10,018 | -42% | 1 | 1 | 0% | 2,275 | 3,078 | +35% | 0 | 0 | — |
case-19 | pass→pass | 16,929 | 11,121 | -34% | 1 | 1 | 0% | 1,918 | 2,421 | +26% | 0 | 0 | — |
case-20 | fail→pass | 12,765 | 5,054 | -60% | 1 | 1 | 0% | 1,013 | 2,092 | +107% | 0 | 0 | — |
case-22 | pass→pass | 17,606 | 16,140 | -8% | 1 | 1 | 0% | 2,185 | 3,179 | +45% | 0 | 0 | — |
case-23 | fail→fail | 26,158 | 18,852 | -28% | 1 | 1 | 0% | 3,399 | 4,266 | +26% | 0 | 0 | — |
case-24 | fail→fail | 27,378 | 25,373 | -7% | 1 | 1 | 0% | 3,880 | 5,188 | +34% | 0 | 0 | — |
case-25 | fail→fail | 25,894 | 22,649 | -13% | 1 | 1 | 0% | 3,749 | 4,689 | +25% | 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. 25 cases were attempted. The headline lift of +20 percentage points is the difference between those two pass rates over the 25 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.