FolderAI

FolderAI · the mass file editor

Hire an AI that already knows the job.

Describe the role in plain words. FolderAI writes the whole document set — who it is, what's off-limits, how it works, which tools it can touch — as a portable folder any model can run. For your assistant, your best employee, a department, or the whole company.

Build mine free →See what you get ↓

No setup. No SDK. Plug the folder into Claude, GPT, or Gemini.

A chat thread isn't an employee.

  • You re-explain the role every Monday. Context dies the moment the tab closes.
  • No two teammates get the same agent — everyone keeps their own private prompt.
  • When the model changes, the behavior drifts and nobody can say why.

What you actually get

Not a prompt. A folder of documents — versioned, portable, readable by a human or a model.

folder/support-lead-agent
├ core/
│  ├ SOUL.md          ← who it is: voice, tone, temperament
│  ├ IDENTITY.md      ← what it will & won't do
│  └ rules.md         ← hard constraints
├ workflows/
│  ├ triage.md        ← how it routes a new ticket
│  └ escalation.md    ← when to hand off to a human
├ tools/
│  ├ helpdesk.md      ← the systems it may touch
│  └ knowledge.md     ← where it looks things up
└ agents/
   └ reviewer.md      ← checks replies before they send

core/SOUL.md

Who the agent is — voice, tone, temperament. Survives a model swap.

core/IDENTITY.md

What it will and won't do — the guardrails, in writing.

workflows/

Step-by-step processes: how it handles each real situation.

tools/

Which systems it can touch, and exactly how — nothing it shouldn't.

How it works

01

Describe the role

Answer a short quiz in plain language. No prompt-engineering required.

02

The AI clarifies

It asks the right questions and proposes answers — one click to accept, edit if needed.

03

Get the document set

FolderAI writes the whole folder — SOUL, IDENTITY, workflows, tools — and checks it for contradictions.

04

Download or refine

One-click ZIP — plug it into your model. Or refine by chat and it edits the files in place.

From one assistant to a whole company

The same document discipline scales. Start with one colleague; grow to an org of them.

Personal assistant

01

Solo founders, operators, creators

One reliable colleague with a SOUL.md and IDENTITY.md that shows up Monday knowing the job — no refresher.

Your best employee

02

Teams hiring their first real AI role

Add memory, tools, and a trust ladder — read-only to autonomous — so it earns responsibility step by step.

A whole department

03

Ops, support, content, data leads

Per-team specs with acceptance tests that travel, so quality is provable — not vibes.

The whole company

04

CEOs & COOs under 'do AI' pressure

A role-by-role audit and a 90-day rollout with cost ceilings — a board-ready plan, not a vision deck.

Built for both sides of the room

If you've never written a prompt

You answer questions in plain language and get a finished, organized folder. No syntax, no SDK, no blank page. If you can brief a new hire, you can do this.

If you live in the terminal

Portable Markdown + JSON you can read, diff, and version. Mass-edit dozens of files at once, gate with eval cases, and plug into Claude Code, the Agents SDK, or your own pipeline.

Why it holds

Specs over vibes

Behavior lives in files you can read and review — not a chat nobody can reproduce.

Portable, not locked in

It's your folder. Plug it into any model, check it into git, hand it to a teammate.

Proven on itself

The same SOUL.md / IDENTITY.md pattern every agent inside FolderAI ships with.

Questions, answered

Do I need to be technical?+

No. You answer a plain-language quiz and the AI fills the gaps. Technical folks get clean, diffable Markdown to take further.

What do I get at the end?+

A downloadable folder of documents — SOUL.md, IDENTITY.md, workflows, tools — that any model can run. No SDK, no lock-in.

Which models does it work with?+

Any. The folder is plain Markdown + JSON — plug it into Claude, GPT, Gemini, or Claude Code.

Can it scale to a team or company?+

Yes. Start with one assistant; the same discipline grows to per-team specs and a company-wide audit with a 90-day rollout.

How is this different from a long prompt?+

A prompt is a chat that dies when the tab closes. A folder is a portable, reviewable spec that survives model swaps and teammates.

The stack we build on

Modern tech, models & agents — production-ready

Frontier and open models, agent frameworks, data and infrastructure down to on-prem — what we ship products and AI agents with.

01

Models

Frontier and open-weights, matched to the job and budget

  • ClaudeAnthropic
  • GPTOpenAI
  • GeminiGoogle
  • GrokxAI
  • LlamaMeta
  • MistralMistral AI
  • DeepSeekDeepSeek
  • QwenAlibaba

02

Orchestration

Agents, tools, MCP and pipelines

  • MCPModel Context Protocol
  • LangGraphLangChain
  • LangChainLangChain
  • LlamaIndexLlamaIndex
  • CrewAICrewAI
  • Pydantic AIPydantic
  • AI SDKVercel
  • OllamaOllama

03

Data

Databases, vector search and ML

  • PostgresPostgreSQL
  • SupabaseSupabase
  • RedisRedis
  • MilvusZilliz
  • ElasticsearchElastic
  • DuckDBDuckDB
  • PyTorchPyTorch Foundation
  • Hugging FaceHugging Face

04

Infra

Cloud, containers and air-gapped setups

  • DockerDocker
  • KubernetesCNCF
  • TerraformHashiCorp
  • Linuxon-prem · air-gapped
  • CUDANVIDIA
  • NGINXF5
  • VercelVercel
  • RailwayRailway

Trademarks belong to their owners. Not partnerships — just the tools we work with.

Describe the role. Get the folder.

One sentence in, a complete AI colleague out. Build your first one free — no setup, no card.

Build mine free →