---
title: Daita — Your data has a new interface
description: A persistent, read-first data agent for grounded analysis, durable outcomes, and controlled action—plus an intelligent pipeline for trusted data ingestion.
canonical: https://daita-tech.io/
---

# Daita

Your data has a new interface.

Daita is a persistent data agent that investigates, packages, and repeats trusted work, plus an intelligent pipeline that turns raw files into production-ready data.

[View the visual website](https://daita-tech.io/)

## 01 / Terminal agent

The agent lives where your work already happens.

Launch Daita and work across live databases, a bounded local workspace, and explicitly admitted remote MCP tools. It grounds every result in inspectable evidence, creates useful deliverables, and carries bounded work forward.

### Capabilities

- **Three governed surfaces:** Query cataloged SQLite and PostgreSQL databases, search and analyze local files, and call exact admitted read-only MCP tools.
- **Artifacts, not just answers:** Create model-authored documents, canonical JSON snapshots, and exact CSV or XLSX exports with retained provenance.
- **Durable work:** Run bounded data-profile jobs and one-time, interval, or calendar read routines with local conversation-inbox delivery.
- **Read first, not answer only:** Analysis starts read-first. Exact file replacement and scoped PostgreSQL updates require explicit configuration, preview, and approval.
- **Context that compounds:** Conversations persist. Memory, business semantics, and reusable skills remain inspectable, advisory, and explicitly managed.

### Operating surfaces

- **Interactive TUI:** Conversations, sources, files, MCP, jobs, routines, inbox, artifacts, and approvals.
- **Headless CLI:** Automation-friendly commands with JSON output.
- **Python API:** Embed the same persistent agent and capability boundaries in local applications.
- **Resident host:** Explicitly keep an agent open so admitted jobs and schedules can progress after the TUI closes.

### Model routes

Use direct OpenAI, Anthropic, Gemini, or Grok APIs; an OpenAI-compatible endpoint; local Ollama models; or supported Codex/ChatGPT, Claude Code, and Grok Build subscriptions.

### Defined boundaries

- No arbitrary SQL writes, inserts, deletes, DDL, or role changes.
- No shell access or generic file writer.
- Remote MCP tools and scheduled routines remain read-only.
- Scheduled outcomes stay in the local conversation inbox; schedules require an open agent host.
- Durable knowledge does not change silently in the background.

### Install

    curl -fsSL https://daita-tech.io/install.sh | bash
    daita

[View Daita Agents on GitHub](https://github.com/Daita-Corp/daita-agents)

## 02 / Agentic pipeline

The agentic pipeline is in development and coming to the Daita WebUI.

Upload a file, select a destination, and watch the pipeline map, validate, transform, resolve, and load your data—asking for context only when it needs you.

1. **Map:** Match source fields to the live destination schema.
2. **Validate:** Check types, constraints, required values, and relationships.
3. **Clarify:** Ask for human judgment when meaning is ambiguous.
4. **Transform:** Generate and apply the required cleanup recipes.
5. **Resolve:** Connect foreign keys and preserve parent-child records.
6. **Generate:** Build correlated SQL only after confidence clears the threshold.
7. **Load:** Execute with a live, inspectable progress stream in the WebUI.

## One system / two directions

- **Ask:** The terminal agent turns your existing data into a persistent, grounded conversation.
- **Move:** The pipeline turns incoming data into validated, connected records inside your destination schema.

## Links

- [Open the Daita WebUI](https://daita-tech.io/app/dashboard)
- [Documentation](https://docs.daita-tech.io)
- [GitHub](https://github.com/Daita-Corp/daita-agents)
- [Contact](https://daita-tech.io/contact/sales)
- [Privacy](https://daita-tech.io/legal/privacy)
- [Terms](https://daita-tech.io/legal/terms)