How do you build a Lindy AI assistant in 2026?
Quick Answer: As of April 2026, building a Lindy AI assistant takes three steps: pick a trigger (calendar event, email, Slack message, or webhook), select skills (Gmail, Calendar, CRM, web search) the agent can use, and write a prompt describing the task. Lindy supports human-in-the-loop confirmation before sending replies or taking external actions.
Building a Lindy AI Assistant
Lindy is a no-code platform for personal AI assistants triggered by everyday events. The build process is intentionally short.
Step 1 — Pick a Trigger
In the Lindy dashboard, click "New Lindy" and select a trigger from the catalog:
- Email received — Gmail or Outlook inbox
- Calendar event — N minutes before a meeting
- Slack message — Direct message or channel mention
- Webhook — Custom HTTP endpoint
- Manual — User-initiated chat
Step 2 — Add Skills
Skills are tools the agent can call. Common skills:
- Send email — Compose and send via the connected inbox
- Schedule meeting — Read calendar availability and book
- Search the web — Bing or Google search
- Update CRM — HubSpot, Salesforce, or Pipedream actions
- Custom action — Webhook with templated body
Lindy enforces OAuth scopes per skill, so the agent only has the access granted at connection time.
Step 3 — Write the Prompt
Lindy uses a system-prompt-style instruction. Concrete prompt patterns work better than abstract ones:
"When a meeting starts in 15 minutes, search the attendee's LinkedIn and recent emails. Summarize key facts in a Slack DM to me. Do not send anything to the attendee."
Step 4 — Review and Approve
Enable "Confirm before sending" so the agent drafts replies and asks for approval via Slack or web before any external action. Once trust is established, switch specific skills to fully automated. Templates from the Lindy library cover meeting prep, inbox triage, and lead enrichment.