The tools
Four AI teams to try
Each one is a real multi-agent system built with LangGraph.js. Pick one, give it a topic or task, and watch the team work.
Article Squad
Nine agents pick a trending topic, plan, vote on the title, draft, argue and approve a 300-word article.
- Plan, workflow and who approves what
- Task board, Draft 1 vs Draft 2
Blog Squad
The same team writes a short post at a round table. Every message pops up so you can follow who talks to whom.
- Round table with roles and power
- Coach chatbot and Learner's Guide
AgentSquad
A leader splits your task across specialists who work in parallel; a critic reviews; the leader merges one answer.
- Startup Founding Team or General Squad
- Go / Pivot / Stop verdicts
AI Startup Idea Lab
A CEO and five specialists brainstorm on a shared whiteboard, help each other, score ideas and draw a mind map.
- Help round between agents
- Scores, mind map and evidence
Learn
How a team of AI agents works
One AI model, many job descriptions. The code decides who speaks, when and to whom, and which agent has the power to decide.
- 1TopicYou give one, or the Researcher scouts trending ideas and the Team Lead picks.
- 2PlanThe Team Lead sets the goal, the quality bar and who joins, and assigns tasks.
- 3VoteThree agents vote on headlines. The winner is locked as the title.
- 4Draft ⇄ reviewThe Editor scores, the Writer revises or pushes back, up to 2 rounds.
- 5ApproveThe Team Lead approves at the quality gate and publishes.
Hub and spokes: most messages go through the Team Lead; some go straight between teammates as handoffs.
A draft passes only when the Editor's score reaches the bar and the Fact-checker has no flags. Otherwise it goes back.
After the brief, the agents decide everything. You watch every message, task and decision, then read how it's built.
Bring your own key
Free with your own Groq, Gemini or OpenRouter key
Every tool runs on your own free key. Get one in a minute from Groq, Google AI Studio or OpenRouter. Choose "Auto" and the team picks the best working model and switches by itself if one is retired or busy.
Groq
Fastest replies. Open models: llama-3.3-70b-versatile, openai/gpt-oss-120b, openai/gpt-oss-20b, llama-3.1-8b-instant.
Google Gemini
Gemini Flash models: gemini-3.6-flash, gemini-3.5-flash-lite, gemini-3.8-flash, gemini-3.1-flash-lite.
OpenRouter
Many free models from different labs (ids ending in :free). The list is loaded live, because free models change often.
In each tool, open "Use your own free API key", pick a provider, paste your key and press Check key and load models to see exactly which models your key can use.
Privacy
Your data, kept out of the way
No sign-up, no tracking cookies, no ads.
The site runs in your browser. Your key goes straight to the provider and never reaches worksmarto.
There is no worksmarto server or database. Topics and answers stay in your browser tab.
India DPDP Act 2023 rights and a grievance contact. Read the privacy policy.
Questions
FAQ
Is it really free?
Yes. The site is free and runs on your own free key from Groq, Google AI Studio or OpenRouter. Your provider's own free-tier limits apply. You can also replay the built-in samples without any key.
Is the output reliable?
It's AI-generated and can be wrong. The team checks itself (Editor, Reader, Fact-checker), but you should still check facts before you use or publish anything. It's not financial, legal or medical advice.
What if a model stops working?
With "Auto", the team tries the next model from the same provider when one is retired, missing or busy, using the provider's live model list.
Can I build this myself?
Yes. Each tool has a Learner's Guide with the agent prompts and the LangGraph.js graph. The code is open source under the MIT licence.
Who made this?
The Worksmarto Team at Software Innovation Labs, Erode, Tamil Nadu, as part of worksmarto.com.