# docbank

> Keep, find, update, and recover documents in a vault you control.

Every public page is also published as raw Markdown at the route listed below.

## Public Site
- [Docbank](https://docbank.ai/index.md): Keep and recover documents for people, applications, and agents.
- [Document guide](https://docbank.ai/guide.md): Follow a document from import through saved versions, processing choices, and recovery.

## Overview
- [Docbank documentation](https://docbank.ai/docs/index.md): Install Docbank, then find the guide for importing, searching, exporting, automating, or backing up a vault.

## Changelog
- [Changelog](https://docbank.ai/docs/changelog.md): Release history.

## Start here
- [Setup](https://docbank.ai/docs/setup.md): Install Docbank on Linux, macOS, or Windows and create the vault.
- [Quickstart](https://docbank.ai/docs/quickstart.md): A ten-minute tour of the docbank CLI.
- [Capabilities](https://docbank.ai/docs/capabilities.md): What Docbank can do today, grouped by task, with links to the guide for each.
- [Visual tour](https://docbank.ai/docs/tour.md): Screenshots of the Docbank web app and terminal browser, taken against synthetic vaults.

## Work with documents
- [Importing documents](https://docbank.ai/docs/usage/importing.md): Import folders, preview large sources, retry partial imports, and keep changing files up to date.
- [Photo assets](https://docbank.ai/docs/usage/photos.md): Group camera files, browse photos in the web app, and organize albums through the CLI or HTTP API.
- [People](https://docbank.ai/docs/usage/people.md): List and edit canonical people through the daemon API and CLI, and read them through MCP.
- [Email PDFs](https://docbank.ai/docs/usage/email-pdf.md): Render one email version as a verified PDF kept alongside its original EML.
- [Organizing and tagging](https://docbank.ai/docs/usage/organizing.md): Browse, move, rename, and tag documents in the virtual tree.
- [Searching](https://docbank.ai/docs/usage/searching.md): Ranked, prefix-matching search over document names and verified text content.
- [Search exports](https://docbank.ai/docs/usage/search-exports.md): Export search counts for a date range as CSV, review date evidence, and verify the frozen evidence ZIP.
- [Document processing](https://docbank.ai/docs/usage/document-processing.md): Preview, consent to, and run document processing. The original version is never changed.
- [Document processing search](https://docbank.ai/docs/usage/search.md): Search only the document versions a consumer currently authorizes.
- [Web application](https://docbank.ai/docs/usage/web.md): Find, read, organize, and export documents in the local, authenticated web app.
- [Interactive terminal browser](https://docbank.ai/docs/usage/tui.md): Browse documents, inspect their details, storage, backups, and permanent history, and move documents to or from trash in the terminal.

## Protect and operate
- [Vault lifecycle](https://docbank.ai/docs/usage/lifecycle.md): Operate a Docbank vault safely from first import through maintenance, upgrades, snapshots, and recovery.
- [Permanent audited history](https://docbank.ai/docs/usage/audited-history.md): Keep every version and recorded change under a directory permanently, after reviewing what the scope covers.
- [Trash, garbage collection, repack, and verify](https://docbank.ai/docs/usage/trash-and-gc.md): How deletion and space reclamation work, from trash through garbage collection and repack.
- [Multi-store storage](https://docbank.ai/docs/usage/storage.md): Keep document bytes local first, add filesystem or S3-compatible stores, and move verified content between them.
- [Backup and restore](https://docbank.ai/docs/usage/backup.md): Create incremental snapshots in a backup repository, verify them, and restore a vault from them.
- [Verify transfer packages](https://docbank.ai/docs/usage/transfers.md): Check a Msgvault transfer package locally before importing it.
- [Verified export bundles](https://docbank.ai/docs/usage/export-bundles.md): Download document versions, verified email PDFs, and attachment sets as verified ZIP bundles.
- [Troubleshooting](https://docbank.ai/docs/troubleshooting.md): Diagnose Docbank startup, daemon, import, integrity, update, and HTTP API failures without risking the vault.

## Automate and integrate
- [Docbank for agents](https://docbank.ai/docs/agents.md): Why agents use Docbank, which interface to choose, and how it keeps document automation safe.
- [Agent integration guide](https://docbank.ai/docs/agents/integration.md): Connect an agent to Docbank using its OpenAPI contract, authenticated HTTP API, revisions, and dry-run maintenance operations.
- [Model Context Protocol](https://docbank.ai/docs/usage/mcp.md): Connect a local MCP client to Docbank's read-mostly document tools through the daemon.
- [Embed in Go](https://docbank.ai/docs/embedding.md): Run one or more Docbank vaults inside a Go application, each with its own root, using CGO or pure-Go SQLite.
- [Document understanding in Go](https://docbank.ai/docs/document-understanding.md): Choose Go packages for document extraction, canonical evidence, renditions, and embeddings without opening a Docbank vault.

## Reference
- [CLI reference](https://docbank.ai/docs/cli-reference.md): Every docbank command, with its flags, output formats, and error behavior.
- [Configuration](https://docbank.ai/docs/configuration.md): Vault location, data layout, config.toml, and environment variables.
- [Document processing configuration](https://docbank.ai/docs/usage/configuration.md): Configure processing profiles the daemon can run, with credentials kept outside the portable policy.
- [HTTP API](https://docbank.ai/docs/architecture/http-api.md): The agent-first HTTP API: filesystem-shaped endpoints, revision preconditions, and the daemon's error contract.

## How it works
- [How Docbank works](https://docbank.ai/docs/architecture/overview.md): A guided model of vaults, document identity, immutable content, storage, deletion, and recovery.
- [Document processing](https://docbank.ai/docs/architecture/document-processing.md): How Docbank derives readable text and search data while keeping original versions and requiring consent for processing.
- [Source metadata](https://docbank.ai/docs/architecture/source-metadata.md): How Docbank records a limited set of facts found inside original files.
- [Format coverage](https://docbank.ai/docs/architecture/format-coverage.md): How Docbank reports detection, retention, metadata, expansion, text, page, and transcript support from the running binary.
- [Document timeline](https://docbank.ai/docs/architecture/document-timeline.md): How Docbank derives date claims from stored document evidence.
- [Visual previews](https://docbank.ai/docs/architecture/visual-previews.md): How Docbank identifies and stores canonical visual derivatives.
- [Verified page images](https://docbank.ai/docs/architecture/page-images.md): Page geometry and local page rendering for one document version, within fixed limits.
- [Storage](https://docbank.ai/docs/architecture/storage.md): The SQLite schema, blob store layout, durability discipline, and enforced invariants.
- [Loose and packed content](https://docbank.ai/docs/architecture/packed-storage.md): The shared Kit packed content layer, and what Docbank itself remains responsible for.
- [Editing and versions](https://docbank.ai/docs/architecture/editing-and-versions.md): Stable content-version identity, retrieval, replacement, reversion, and retention over immutable content.
- [Ownership and concurrency](https://docbank.ai/docs/architecture/locking.md): How daemon and embedded owners coordinate concurrent access with SQLite, hierarchy locks, and owner-local operation gates.
- [Daemon and process model](https://docbank.ai/docs/architecture/daemon.md): The single process that owns the vault, docbank daemon run, and how the CLI discovers, auto-starts, and stops it.
- [Backup and recovery](https://docbank.ai/docs/architecture/backup.md): How Docbank snapshots and restores a vault using JSONL metadata and a Kit repository.
- [Integrity and trust](https://docbank.ai/docs/architecture/integrity.md): What Docbank defends against, which layer owns each integrity guarantee, and the trade-offs that were considered and not taken.
- [Audited history](https://docbank.ai/docs/architecture/audited-history.md): The permanent, tamper-evident history model for protected directory scopes, content versions, backup, agents, and future interactive clients.
- [Document derivatives](https://docbank.ai/docs/architecture/document-derivatives.md): Authority, retention, retrieval, and rendition format for processed document derivatives.

## Project
- [Roadmap](https://docbank.ai/docs/roadmap.md): What Docbank does today, what it does not do yet, and what is planned.
- [License](https://docbank.ai/docs/license.md): Docbank is open-source software licensed under the Apache License, Version 2.0.