Snapi Documentation
Snapi is an AI work intake and execution platform — it connects embeddable forms, kanban boards, webhooks, and AI agents in one place so that scattered requests become finished work.
What is Snapi?
Most teams receive work through a mix of emails, Slack messages, spreadsheets, and ad-hoc requests. Snapi gives you a structured way to capture that intake — through forms you design — and routes it through a workflow you control, with AI agents ready to act at every step.
The core idea is simple: define a form, embed it anywhere, and let submissions flow into your board. From there, webhooks, API keys, and an MCP server let your AI agents read, respond to, and complete the work.
Key concepts
| Concept | What it is |
|---|---|
| Form | A form definition — its fields, validation rules, and webhook URLs. Think of it as a table schema for a type of request. |
| Record | One filled-in submission of a form. Each record stores the field values, timestamps, and any attached files. |
| Board | A Kanban board with columns you define. Any record can be added to any board as a card. |
| Webhook | An HTTP callback — pre-webhooks fire before a record is saved (and can abort it); post-webhooks fire after. |
| API Key | A bearer token your AI agents or external tools use to read and write Snapi data programmatically. |
| MCP | Model Context Protocol server — 33 tools that let AI agents interact with Snapi directly inside their context window. |
How it fits together
Here is a realistic example. You run a software agency and want clients to be able to report bugs without emailing you.
You create a Bug Report form with fields for title, description, priority (a select field), and a screenshot upload. You copy the generated iframe and paste it into your client's support page. Their users fill it out — no Snapi account needed.
Each submission becomes a Record, and immediately appears as a card on your Bug Tracker board in the "New" column. A post-webhook fires to your n8n instance, which labels the card by priority and sends a Slack notification. Your AI agent, connected via MCP, picks up new cards in its next loop — reads the description, checks your codebase, drafts a fix, and posts a comment on the card with its findings.
You review the comment, drag the card to "In Progress", and start implementing. The audit log tracks every step.
No AI agent required. Snapi works just as well without AI — the forms, boards, and webhooks are useful on their own. Agents are an optional layer you add when you're ready.