FTAI - Human-AI Format for VS Code

Full language support for .ftai files — the human-readable, AI-parseable format designed for the age of agents.

FTAI replaces the JSON + Markdown + YAML triad with a single format that's readable by humans, parseable by machines, and native to AI agents.

Features

Syntax Highlighting

Rich, semantic highlighting for all .ftai constructs:

@tags — block tags like @document, @section, @schema, @task
@@nested tags — subsections and depth indicators
key: value metadata — clearly distinguished keys and values
Type hints — :int, :float, :bool, :date suffixes
// comments — inline comments (URL-safe)
**bold** emphasis
@table blocks with column separators
Quoted strings, numbers, booleans

Document Outline

Navigate .ftai files with the Outline panel and breadcrumbs:

Every @tag appears as a navigable symbol
Hierarchical structure (@section → @@subsection) is preserved
Tag types get distinct icons (sections, tables, tasks, tool calls, memory blocks)

Live Validation

Real-time diagnostics as you type:

Missing @ftai v2.0 header detection
Missing required @document block
Unclosed @end block warnings
Unexpected @end without matching open block
Run manually via Command Palette: FTAI: Validate Document

Snippets

Quick-start templates for common .ftai blocks:

Prefix	Block
`ftai`	New document with header and metadata
`section`	`@section` block
`subsection`	`@@subsection` block
`schema`	`@schema` with required/optional tags
`table`	`@table` with header and divider
`task`	`@task` with status and priority
`tool_call`	Tool invocation block
`memory`	Memory block with scope selection
`ai_note`	AI annotation block
`goal`	Single-line goal
`constraints`	Constraints list
`include`	Include another `.ftai` file
`image`	Image reference block

Code Folding

Fold blocks between @tag and @end for easy navigation of large documents.

Example

@ftai v2.0

@document Cardiac Arrest – Adult Algorithm
  owner: Protocol Committee
  date: 2025-04-28
  version: 1.2.0
  memory_scope: long_term
@end

@section Cardiac Arrest – Adult

  @@subsection Initial Actions
    - **Start high-quality CPR 30:2**  // chest depth 5-6 cm
    - Attach AED/monitor (interrupt ≤10 s)

  @@subsection Drug Reference
    @table Drug Doses
      Name | Route | Adult | Pediatric
      ---  | ---   | ---   | ---
      Epinephrine | IV/IO | 1 mg q3-5 min | 0.01 mg/kg
      Amiodarone  | IV/IO | 300→150 mg | 5 mg/kg once
    @end
@end

@goal Provide deterministic resuscitation steps.

@result Patient achieved ROSC after 2 shocks.

What is FTAI?

FTAI is an open-source format (Apache 2.0) created by FolkTech AI for human-AI collaboration. It's designed to be:

Readable at a glance — no curly braces, no escape characters
Flexible by design — prompts, protocols, agent memory, SOPs, study materials
Safe and stable — no indentation crashes, no injection risks
Agent-native — AI-to-AI communication in a standardized format

FTAI is used in production across multiple products including AI assistants, medical protocol apps, and educational study platforms — with 868+ .ftai files in active use.

Requirements

VS Code 1.85.0 or later
Also works with Cursor

License

Apache 2.0 — FolkTech AI

FTAI - Human-AI Format

Folktech AI