> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tight.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AI-Native GL

> Leveraging data from over 1.3M SMBs and a fully deterministic General Ledger

## Best of everything AI-native GL (AGL)

A fully deterministic general ledger ensures accuracy and reliability, but traditional ledgers weren't built to
take full advantage of AI. Tight built its AI-native General Ledger to both take advantage of the quick evolution of
LLMs and leverage the accuracy of a fully deterministic ledger.

### What AI can automate

AI helps automate categorization (93% on average), reconciliation, anomaly detection, reporting, and analysis, while the
fully deterministic nature of Tight's GL ensures AI hallucinations don't impact the underlying GL.

### Why a deterministic General Ledger

By nature of their model, LLMs work best with human language-level abstractions, not writing debits and credits
directly into the GL. Tight's API provides these LLM-friendly abstractions, e.g. categorizing a transaction, marking a
bill as paid, refunding an invoice payment, while encapsulating the underlying writes into the GL in a proven, robust
architecture. This API-first design is the *perfect* use case for Agents to fully manage a user's books without needing
to manage low-level general ledger transactions directly.

## Agentic Accounting

Agentic Accountants have arrived, allowing business owners to keep their books on auto-drive while trusting that Agentic
Accountants will reach out with any questions if/when needed. See
Tight's [Agentic Accounting updates](/ai/agentic/agentic-accountant) to stay up to date on its rapid
evolution.

## Creating an AI-native architecture

Tight's most talented (both experienced and AI-native) engineers reenvisioned Tight's APIs and architecture with Agentic
Accountants in mind, yielding Tight's AI-native ledger:

* [v6 API](/api-reference/authentication)
  * Proven maturity and edge-case handling from v5
  * DB optimized for high-performance floor and ceiling for AI-centric use cases, including burst requests
  * Result sets formatted for ease of consumption by LLMs
* [Self-learning transaction categorization and reconciliation](ai/categorization/self-learning-transaction-categorization-overview),
  leveraging rules formed by 1.3M SMBs' worth of data

## Building Tight MCPs

The [Tight MCP Server](/ai/mcp-tools/mcp-overview) was designed in parallel with the Tight v6 API itself, naturally
creating a cohesive AI-native ecosystem. That combined with the Tight team's deep knowledge of accounting use-cases,
both for business owners and bookkeepers, translated to a comprehensive set of MCP tools enabling agentic interaction
with the Tight API.

The Tight API's MCP tools are grouped by domain area:

* [Bank Transactions](/ai/mcp-tools/bank-transactions-mcp)
* [Embedding & Integration](/ai/mcp-tools/embedding-and-integration-mcp)
* [Financial Statements](/ai/mcp-tools/financial-statements-mcp)
* [Invoicing](/ai/mcp-tools/invoicing-mcp)

## Next Steps

<CardGroup cols={2}>
  <Card title="MCP Tools" icon="rocket-launch" href="/ai/mcp-tools/mcp-overview">
    Integrate Tight's rich accounting knowledge into your LLM-powered product
  </Card>

  <Card title="Agentic Accounting" icon="sparkles" href="/guides/launch-paths/embed-in-your-ux">
    Enable autonomous AI agents to handle end-to-end accounting workflows
  </Card>
</CardGroup>