20 min free · no signup · real results
🛡 Security Audited 🔒 Zero Password Sharing

Assign a Task.
Get It Done.

AI professionals that don't just advise — we deliver. Each one is a digital clone of a real specialist. Describe your task, get a real result. Not a conversation — a deliverable.

6
Specializations
50+
Years Combined
20 min
Free per Doer
24/7
Availability

Chat free. Pay only for work.

Every doer has a personal assistant that handles initial communication — for free. The billable timer starts only when your task is handed off to the specialist.

1
Free

Chat with Assistant

Describe your task in any detail. The assistant validates relevance, asks clarifying questions, and builds a clear spec. This phase is always free.

"I need a data pipeline built" → "What source? What volume? Where does it land?"
2
Handoff

Task Assigned to Specialist

Once the assistant has enough info — data sources, requirements, credentials — it hands the task to the doer. The billable session starts here.

"Task ready: Build ETL from S3 CSV to BigQuery with dbt. Timer starts now."
3
Billable

Doer Delivers

The specialist works with real tools — generates code, reports, configs. You see real-time progress. When done, the assistant summarizes the result.

✓ Architecture diagram + dbt project + Airflow DAGs

Pick your specialist

Each doer is a digital clone of a real professional. Same skills, same methodology — but available 24/7 and works in minutes.

📊

Data Engineer

Pipelines, Architecture & Analytics
👤 clone of Vladimir Kobzev

Builds data pipelines, designs warehouse architectures, writes complex SQL. From raw data to production-ready analytics.

Example tasks
Design a data warehouse schema
Build an ETL pipeline with dbt
Optimize slow BigQuery queries
SQL Python dbt BigQuery Airflow Spark
🎨

Digital Designer

UI/UX, Design Systems & Prototyping
👤 clone of a senior designer

Creates wireframes, reviews interfaces, builds design systems. Thinks in components, delivers in specs and code.

Example tasks
Audit a landing page UX
Create a design system spec
Redesign a checkout flow
Figma Design Systems UX Audit Prototyping
📈

Investment Analyst

Financial Modeling & Market Research
👤 clone of a senior analyst

Builds financial models, runs DCF valuations, analyzes markets. Delivers spreadsheets, memos, and investment theses.

Example tasks
Build a DCF model for a SaaS company
Analyze a market segment
Write an investment memo
DCF Comps Market Analysis Excel
✍️

Content Marketer

Strategy, SEO & Copywriting
👤 clone of Ilya Telegin

Writes blog posts, builds content strategies, does SEO audits. Delivers publish-ready articles, calendars, and distribution plans.

Example tasks
Write a 2000-word blog post
Create a 3-month content calendar
SEO audit for a landing page
Content Strategy SEO Copywriting Social
⚙️

Simulation Engineer

Injection Molding & CAE Analysis
👤 clone of a CAE specialist

Analyzes mold designs, troubleshoots defects, selects materials. Delivers simulation reports, DFM reviews, and process parameters.

Example tasks
Review a part design for moldability
Troubleshoot warpage defects
Select material for automotive part
Moldflow DFM Materials FEA
💻

DevOps Engineer

Infrastructure, CI/CD & Cloud
👤 clone of a senior SRE

Sets up infrastructure, writes CI/CD pipelines, debugs deployments. Delivers Terraform configs, Dockerfiles, and runbooks.

Example tasks
Set up a Kubernetes cluster on GCP
Write a GitHub Actions CI pipeline
Dockerize a Node.js + Postgres app
Docker K8s Terraform AWS GCP CI/CD

Doers, not chatbots

ChatGPT gives advice. AI Doer gives deliverables.

📦

Deliverables, Not Advice

You get code, configs, reports, spreadsheets — actual work product. Not "here are 5 things to consider" but a ready-to-use artifact.

🔌

Real Tools via MCP

Each doer is connected to professional tools through Model Context Protocol. They query databases, run simulations, generate diagrams — not just text.

🧠

Cloned from Real Pros

Not a generic model with a system prompt. Each doer is built from a real professional's methodology, knowledge base, and decision-making patterns.

🔒

Your Secrets Stay Yours

API keys, credentials, tokens — stored in a secure vault. The AI doer uses tools through MCP without ever seeing your raw secrets.

ChatGPT / Claude Freelancer AI Doer
Output type Text advice Deliverables Deliverables
Domain expertise Generic Deep Deep (cloned)
Uses real tools No Yes Yes (MCP)
Turnaround time Instant Days Minutes
Available 24/7 Yes No Yes
Cost for 1 task ~$0.05 $50-500 $0 (free trial)

See a doer in action

A real task assigned to the Data Engineer

DE

Data Engineer

Working on your task
16:28 remaining
Y
I have 50GB of CSV logs landing daily in S3. Build me a pipeline to BigQuery for analytics. Budget-friendly, batch is fine.
DE
Got it. Working on your pipeline now.

Generating architecture diagram...
DE
Done. Here's what I've built for you:

1. Architecture: S3 → SQS trigger → Spark (EMR Serverless, daily batch) → BigQuery
2. Transform layer: dbt with medallion pattern (bronze/silver/gold)
3. Orchestration: Airflow DAG with 3 tasks, daily at 02:00 UTC

Estimated cost: ~$12/day on EMR Serverless for 50GB batch.

✓ 4 files generated
DE
Deliverables ready:
📄 architecture.mermaid — pipeline diagram
📄 dbt_project/ — full dbt scaffold (models, sources, tests)
📄 airflow_dag.py — orchestration DAG
📄 terraform/bigquery.tf — warehouse infra

Want me to adjust anything or add monitoring?

Security-first. Audited. Open source.

Your credentials never touch the AI. Full isolation, zero password sharing, verifiable by design.

🛡

AI doers have zero access to your keys

When you store a credential in the vault, it's encrypted and accessible only through MCP tools — never through the AI conversation. The doer can use a tool that needs your API key, but it never sees the key itself. When you revoke access, it's gone instantly.

✓ Security Audited ✓ Zero Password Sharing ✓ Open Source ✓ Full Session Isolation

🔒 Vault, Not Chat

Credentials are stored in World MCP vault, an encrypted store separate from the AI context. The doer calls tools through MCP protocol — raw secrets never enter the conversation, the prompt, or the AI's memory.

🚫 No Password Sharing

You never type a password into the chat. Credentials are entered on a separate secure page, stored encrypted, and passed directly to tools. The AI agent has no mechanism to read, log, or exfiltrate them.

🧩 Full Session Isolation

Each task runs in a completely isolated session. No shared memory between tasks, no cross-session data leakage, no persistent state. When a session ends, everything is wiped. One user's data never touches another's.

👁 You Control Access

Each credential has explicit permissions. You decide which tools can use which keys, and you can revoke access at any time. When access is revoked, the doer loses it immediately — no cached copies, no delayed revocation.

🛠

Fully open source and auditable. The World MCP vault is an open-source project — inspect the code yourself, verify there are no backdoors, and confirm the isolation model. Security through transparency, not obscurity. View on GitHub →

How doers work

Each doer is a Claude Code agent with domain-specific tools, running on dedicated hardware.

You
🌐 ai-doer.chillai.space
💬 Chat UI (WebSocket)
🔒 Credential Vault
WebSocket / Cloudflare Tunnel
Assistant Layer FREE
💬 GPT-4o-mini (fast, cheap)
🔎 Task validation & triage
📝 Spec builder & clarifier
Handoff → task is ready, timer starts
Doer Layer BILLABLE
💻 MacBook Air (M-series, 8GB)
🤖 Claude Code per task
📄 Expert CLAUDE.md + KB
MCP Protocol
Tools
⚙ MoldSim MCP
🌐 Playwright MCP
📊 Analytics MCP
🔒 Vault MCP (secrets)
Two-tier architecture: When you open a chat, you first talk to a lightweight assistant (GPT-4o-mini) — fast, free, and always available. It validates your task, asks clarifying questions, and gathers requirements. Once the spec is ready, the assistant hands off to a dedicated Claude Code session on Apple Silicon hardware. Only this deep-work phase counts against your timer. Your credentials are stored encrypted in the Vault MCP — the doer uses them via API proxy and never sees raw secrets.

Start with a free task

20 minutes free with any doer. No signup, no credit card.

Free
$0
Try any doer. 20 minutes of specialist work per doer.
  • Unlimited assistant chat (free)
  • 20 min specialist work per doer
  • Access to all 6 doers
  • Full tool usage
  • No signup required
Try Free
Team
$199/month
For teams that need AI specialists at scale.
  • Up to 10 team members
  • Shared credential vault
  • Custom doer training
  • Admin dashboard
  • Team task history
  • Slack integration
Contact Us

Questions

An AI Doer is a Claude Code agent configured with a real professional's knowledge, methodology, and domain-specific tools. Unlike a generic chatbot that gives advice, a doer executes tasks and delivers work product — code, reports, configs, analyses. Each doer is a digital clone modeled after how a specific specialist works.
Anything within the doer's domain. The Data Engineer builds pipelines and SQL queries. The Designer audits UIs and creates specs. The DevOps Engineer writes Terraform and CI/CD configs. Think of it as having a specialist on call who works in real-time and delivers files, not just text.
Your credentials are stored in a secure vault on the platform. The doer accesses them through MCP tools — it never sees raw passwords, tokens, or API keys. Secrets are passed directly to the tools that need them, completely bypassing the AI conversation.
Three things: (1) Deliverables — you get actual files and artifacts, not text suggestions. (2) Tools — doers connect to real professional tools via MCP, not just language generation. (3) Specialization — each doer follows a real professional's workflow, not a generic assistant pattern. It's the difference between asking someone "how do I build a pipeline?" and saying "build me a pipeline."
Click "Assign Task" on any doer. You'll first chat with a lightweight assistant (always free) that helps clarify your task. Once the spec is clear, the assistant hands off to the specialist — that's when the 20-minute timer starts. You only pay for the specialist's actual work, not the initial conversation. You get 20 minutes per doer, so you can try all 6.
The assistant (GPT-4o-mini) is fast and cheap — it responds instantly, validates your task, and builds a clear spec. The specialist (Claude Code) is powerful but resource-intensive. By separating them, we keep chat free and instant while ensuring the specialist only spends time on well-defined, actionable tasks. You get better results and we don't bill you for "can you clarify?"
Yes, currently each doer runs as a Claude Code session on dedicated Apple Silicon hardware. This gives them access to local MCP tools, file system operations, and full development environments. As we scale, sessions will distribute across multiple hosts, but the architecture stays the same — each task gets its own isolated environment.

Stop asking. Start assigning.

20 minutes free per doer. No signup. Pick a specialist and send your first task now.

Assign a Task