Overview
Respell is a no-code platform for building AI-powered workflow automations (called Spells) by chaining together LLM steps, agents, and integrations. It lets teams automate research, document writing, data entry, and customer outreach without writing code.
Our take
Respell builds AI-first automations through a visual editor, with a free tier that allows genuine evaluation before the jump to the team plan. It sits between traditional automation tools and agent platforms. The part worth attention is the same as everywhere in automation: what happens on the run that fails, and whether anyone finds out.
Worth it if
Teams automating work where inputs vary enough that rule-based tools keep breaking.
Look elsewhere if
Simple, predictable integrations — conventional automation is cheaper and easier to debug.
✨ Features
No-Code Builder
Drag-and-drop interface lets anyone assemble multi-step AI workflows without coding.
AI Agents
Autonomous research and task agents perform web searches and complete multi-step jobs.
Tool Integrations
Connects with spreadsheets, documents, and other apps to read and write data.
Triggers And Scheduling
Spells can run on schedules or be triggered by events to automate recurring work.
⚖️ Pros & Cons
Pros
Cons
💰 Pricing
Free
$0 /mo
- 75 chat messages/month
- 25 spell runs/month
- Access to core agents
Starter
$19.99 /mo
- Billed yearly
- 450 chat messages/month
- 150 spell runs/month
- Research Agent & all integrations
- 1 workspace user
Team
$59.99 /mo
- Everything in Starter
- 1,500 chat messages/month
- 500 spell runs/month
- Unlimited workspace users
- All agents incl. Call Agent
Enterprise
Custom
- At-cost AI usage billing
- Pre-purchase usage at a discount
- Contact sales for details
Categories
❓ Frequently Asked Questions
When does an AI step beat a rule?
When the input varies in ways a rule cannot capture — classifying free-text messages, extracting a value from documents that are laid out differently each time. Where the input is consistent, a deterministic rule is cheaper, faster and fails loudly rather than quietly producing a confident wrong answer.
How do I know an automation has stopped working?
Only if you built that in. Silent failure is the characteristic risk — a token expires, an API changes, a filter stops matching, and nothing happens for weeks. Require a success notification on a schedule and alert on failure, then periodically check the output is correct rather than merely present.