Build powerful automations in Mindsets visual studio
Turn an idea, spec or existing skill.md into agents with tools, if-this-then-that automations and sub agents by speaking to AI
Turn skills on laptops into governed automations - by speaking to our AI
Tell Mindset AI what you want to build or give it your skill.md files & MCPs your team already built and turn them into governed, reliable automations.
- Skills and MCPs Hand over what your team wrote and get back an agent with its functions and connections.
- Approved plans You get a build plan first: agents, functions, connections. Nothing is created until you say yes.
- Reuse first It checks what you already have, so you don't end up with four copies of one connection.
Connect & govern the systems your agent needs to reach
Set up new connections in one place. Every system your agent reaches goes through it, where you can see what it can do, approve it and take it away.
- Connection types Connect to 1000s of agent tools or add your own MCP servers, HTTP APIs, Postgres, Google Sheets, knowledge sources.
- Live-tested setup Our connections builder agent sets up your MCP, APIs & more for you, using documentation, URLs or whatever you provide it with
- Approvals and secrets A person approves anything with side effects, and credentials never reach the model.
Give an agent an if-this-then-that workflow by speaking with AI
Calculations, data transformations, lookups, calls into your systems that stay the same every run. The agent decides when to use it or you force it to use it under specific circumstances.
- Built by AI Describe the automation in natural language, plus the connections it needs. AI builds it for you.
- Deterministic steps Calculations, lookups and transformations run identically every time, and never call the model.
- Model steps Add a model call only where you want judgement, like drafting a reply.
Split the work across specialist agents
A conductor agent delegates to a team of specialists, so complex work can be handled by focused agents that each do one thing well, instead of one over-loaded prompt.
- Any agent can be a resource Attach an agent to another agent the same way you'd attach a function or a connection.
- Focused agents Each specialist can carry simple & short instructions instead of one giant prompt and tool set, increasing performance & reducing LLM costs.
- Permissions A specialist runs with the access of whoever asked so responses are always permission based, and every hand-off is recorded.
Give agents a process that makes it consistent & dependable
Without a script, an agent is a prompt and tools, and you're trusting the model to follow your instructions. It will 'probably' follow them; that's it. A script makes it dependable & cost-effective.
- Phases Break the job into ordered phases, each with a goal and only the tools, agents & functions it needs to complete the job.
- Gates A condition it must pass to move on. Until then it keeps working the user toward it.
- Provable behaviour Write down how it must behave, test it, and guarantee dependability.
Design rich UI your agent shows in chat
Someone asks to book a demo and a calendar appears. Someone asks to change plan and a plan selector renders in the chat.
- Interactive Someone picks a date or a plan in the chat, and the agent acts on it.
- Live data Widgets read from your connected systems, so people see what is true right now.
- Visual guides Or use a card or chart simply to make a long answer readable at a glance.