Build AI agents you can see, test and trust. On a canvas, not in a black box.
Drag agents, knowledge, conditions and human approvals onto one canvas - or ask Vippy to draft the workflow for you. Trigger it from your Apragya apps, a schedule or an API call, keep a person in the loop where it matters, and see what every run did and cost.
Your agents work the night shift
Agents built in Agent Studio run on events, schedules and API calls - and stop for a person when it matters.
- Fire the moment a record is created or its status changes in any Apragya app
- Run on a schedule or when a webhook arrives from another system
- Pause for human approval and resume with the reviewer's decision
- Answer callers through their own API, MCP or chat endpoint - every run metered and logged
Describe the agent in a sentence. Refine it on the canvas.
Ask Vippy for a workflow and it drafts the blocks for you. Then every step is a shape you can open, change, test and run - agents, knowledge, logic, approvals and the systems you connect.
Answers from your documents, on the models you choose.
Upload policies, contracts and process docs into a knowledge base and the agent answers from them - combining meaning search with keyword search. Pick the model per agent from the providers your organisation has keys for; keys stay governed by Apragya, never pasted into a workflow.
- Knowledge bases with semantic + keyword search
- Anthropic, OpenAI, Google, Mistral, Groq, xAI, DeepSeek, Moonshot
- Bring your own keys, or use your organisation's shared key
- Model picker shows only providers you can actually use
- Evaluator block to score an agent's answer
- Guardrails block on input or output
Agents that start themselves - from your apps.
Bind an agent to an Apragya app and it runs the moment a record is created, uploaded, or changes status or a field - with the record mapped in as input. Or run it on a schedule, from a webhook, or on demand.
- Record events: on create, on upload, on status or field change
- Input mapped from the record, scoped to its company
- Schedules and webhooks on the canvas
- App-event trigger block and an Apragya Apps block to read records
- On-demand runs from a button or the API
- Connectors to reach your other systems - including on-prem via the Tenant Runner
From an idea to an agent your team relies on
Describe, design, test, deploy, trigger, review - the same path for a one-step helper or a multi-agent workflow.
A person decides where it matters. Every run on the record.
Drop an approval block anywhere in the workflow. The request lands in your Approvals list, the reviewer approves, rejects or edits the data, and the run picks up where it paused. Every run shows its steps and the AI Credits it used.
- Approval blocks that pause and resume runs
- Reviewers approve, reject or modify in the Approvals list
- AI Credits checked before a model key is handed out
- Credits billed per run, visible in the run logs
- Guardrails block on agent input or output
- Role-based access and an audit trail on the platform
One agent, four ways in.
Deploy a tested workflow as an API with its own keys, as an MCP server other AI tools can call, as an agent-to-agent endpoint, or as a chat. Share it with the rest of your organisation through your private marketplace.
- API deployment with keys you create and revoke
- MCP server and A2A endpoints
- Chat deployment
- Publish privately to your organisation's marketplace
- Send results to Slack, Teams, WhatsApp or email blocks
- Templates to start from
Connected to the tools your team already uses
Agents reach the systems your business runs - through Apragya's governed connector library and the Tenant Runner for on-prem systems, with every call audited.
Every connector is governed: tenant-scoped credentials, an audit log on each call, and the Tenant Runner reaches on-prem systems with no inbound firewall rules. Need one that isn't here? The custom adapter wraps any REST API.
Start with a sentence. Vippy drafts the workflow.
Describe what the agent should do and Vippy lays out the steps on the canvas as a draft - agents, knowledge, approvals and the next action - for you to review and refine.
Free agents, forever. Scale when you're ready.
One subscription per workspace. Includes every Apragya app you turn on - Agent Studio, CRM, HRMS, Finance, and 12 more. Pricing from the live apragya.ai pricing page.
Free Forever
No credit card. No expiry. Keep it as long as you want.
- Free agents - build them, run them forever
- 1,000 AI Credits for building and runs
- Visual canvas with 280+ blocks
- Knowledge bases with semantic + keyword search
- Test runs on the canvas, every run logged
- Community support
Custom
For teams putting AI agents into real business processes. Priced on users, agents and AI usage.
- App-event triggers and scheduled runs at scale
- Human approval steps in your Approvals list
- Deploy as API, MCP, A2A or chat with API keys
- App-event, schedule and webhook triggers
- Private marketplace publishing for your organisation
- From 2,000 to 5,000+ AI Credits / user / month
Every paid plan is tailored to your team, apps and AI usage. Need more apps, custom domains or more AI Credits? Talk to sales →
Honest answers to honest questions
Do I need to code to build an agent?
No. Describe the agent to Vippy and it drafts the workflow on the canvas, or drag blocks yourself. Developers can still call the deployed agent through its API, MCP or A2A endpoint.
Which AI models can agents use?
Anthropic, OpenAI, Google, Mistral, Groq, xAI, DeepSeek and Moonshot - whichever your organisation has keys for. Keys are governed by Apragya, and the model picker only shows providers you can use.
Can a person approve before the agent acts?
Yes. Add an approval block; the request appears in your Approvals list, the reviewer approves, rejects or edits the data, and the run resumes from where it paused.
How do agents connect to our Apragya apps?
Bind an agent to an app and it runs on record events - create, upload, status or field change - with the record as input. Trigger and Apragya Apps blocks on the canvas start runs from app events and read your app records.
Is it really free?
Yes. On the Free Forever plan you build agents and keep running them with no expiry and no credit card, with 1,000 AI Credits. Custom plans add human approval steps, API deployment, triggers and more credits.
How are agent runs billed?
In AI Credits from your organisation's pool. Credits are checked before a run gets a model key, and each run's credits are shown in the run logs.
How is this different from the Web App Builder?
The Web App Builder makes apps people use - screens, forms and records. Agent Studio makes agents and workflows that do the work, on a visual canvas. They work together: an app can run AI on its records, and an agent can be triggered by an app's events.
Put an agent on your busiest process
Free agents, forever. Design it on the canvas and see exactly what every run did and cost.
Start free - no credit card →Not just an agent builder. The whole AI operating system.
Every Apragya subscription unlocks 15+ enterprise apps + Vibe Studio + Agent Studio + Business Suite. Your agents run on the same records, roles, approvals and AI Credits as CRM, HRMS, Finance and the rest.
Vibe Studio
Describe what you want in plain English. Vibe Studio builds the app - screens, data, API, agents - and publishes it to your workspace. No developer required.
Agent Studio
Drag-and-drop canvas to orchestrate AI agents, approvals, humans-in-the-loop. Connect to any system. Governance + audit log built in.