Knowledge & RAG
An agent that guesses is worse than no agent at all
A model with no grounding invents policy, quotes the wrong price, and contradicts your docs — confidently. Pasting knowledge into a prompt doesn't scale, goes stale the moment a doc changes, and blows past the context window.
What's in the box.Each one in the product today.
- 01
Knowledge bases
Create knowledge bases from your documents, scoped per workspace and per agent.
- 02
Chunking & embeddings
Content is chunked and embedded automatically so it's retrievable at conversation time.
- 03
Retrieval (RAG)
The agent retrieves the most relevant chunks per turn and answers from them, with the source in reach.
- 04
Connector sync
Sync knowledge from connectors like Notion; a re-sync keeps the base current as docs change.
- 05
Per-agent scope
Attach a knowledge base to the agents that should use it — nothing bleeds across scope.
Documents are the single grounding surface — chunked, embedded, retrieved
Rather than stuffing a prompt, knowledge is native: ingested from files or connectors, chunked and embedded on the way in, and retrieved per turn at conversation time. Keeping a base current is just a re-sync away.
- Source
- PDF upload
- Base
- Admissions
- Prospectus 2026.pdf—
- chunk—
- embed + index—
- retrieve this turn—
- starting…
Answers pinned to your content, scoped per agent
Each knowledge base is chunked, embedded, and attached to exactly the agents that should use it. At conversation time the agent retrieves the most relevant chunks for the question and answers from them, so responses track your docs — and a re-sync from Notion keeps them current.
- Agent
- Admissions
- Workspace
- Riverside
- Prospectus—
- Fee sheet—
- Staff handbook—
- starting…
Explore the platform.
It works with the rest of Mirakash. Here is what connects to it.
- Agent builderDesign a voice or chat agent with no code — persona, voice, knowledge, tools — and test it live in the browser.
- Workflows & automationAn event-driven, node-graph engine that turns any signal into the right action, automatically.
- Omnichannel & web chatOne agent across phone, WhatsApp, SMS, email, and an embeddable, white-labeled web chat on your domain.
- Use casesA real conversation for six desks, start to finish.
- SecurityRoles, redaction, retention and the audit trail.
- PricingPer minute or per resolution, and what each includes.