Doca is an MIT-licensed, self-hosted document management and knowledge base system, written in TypeScript, that puts rich-text documents, Markdown, spreadsheets, presentations, an infinite canvas and a personal AI assistant into a single workspace for individuals and small teams.

What it is

Doca lives in the TypeScript and React ecosystem and ships as a Docker-deployable web application that a team runs on its own server. Its premise is that five document formats belong in one workbench: rich-text documents, Markdown, online spreadsheets, presentations and an infinite canvas, each with real-time collaboration, comments and document history. Around them sits knowledge management — a knowledge base organised by document catalogue, and collaborative knowledge books (知识册) that support orchestration, source traceability and human review.

The problem it addresses is fragmentation. Project plans, meeting notes, budgets and process descriptions otherwise end up in separate single-format editors, each with its own sharing model, while the reasoning behind a conclusion is lost. Doca places those formats, together with personal files, shared folders, permissions, share links, search, notifications and a trash, behind one account system, and adds a personal AI assistant that works over the documents and files a user is permitted to access. A deployment carries one account system only, and templates, asset providers and optional plugins are installed separately from the default image.

Key capabilities

• Five editors in one workbench — rich-text documents, Markdown, spreadsheets, presentations and infinite canvas — with real-time collaboration, comments and document history.
• A knowledge base with a document catalogue, plus collaborative knowledge books (知识册) with orchestration, source traceability and human review.
• A personal AI assistant that reads PDF and Office attachments, generates and edits images and analyses web material, using a model and service the operator configures.
• A Chrome / Edge browser extension that captures a whole page or selection while preserving body text, source and images, saving into a personal document, a knowledge base folder, or appending to a specified document.
• A Streamable HTTP MCP service that lets MCP-capable clients such as Codex discover authorised tools and SKILL manuals; personal services act as the user, company services name a target user per operation, and tokens can be reset, disabled or revoked.
• Local or S3-compatible file storage alongside share links, search, notifications and a trash.
• Docker deployment on single-node SQLite, or multiple replicas sharing PostgreSQL, Redis and file storage, extended through the public @smartdoca/plugin-sdk.

Who uses it and how

• Individuals and small teams keeping a private workbench for project proposals, meeting notes, budget tracking and process documentation, inviting collaborators to edit in real time and discuss details in comments.
• Teams turning conclusions into durable knowledge, assembling sources, orchestration and human review into knowledge books rather than leaving them in scattered files.
• Self-hosters who want their own AI: deploy on their own server, connect their own model and service, then extend the workbench with business plugins from the plugin store.
• Organisations standardising accounts, using password, OIDC, Google, GitHub, WeChat QR or QQ login with Chinese or English interfaces, depending on what they configure.
• Integrators wiring Doca into external AI clients: individuals enable MCP in personal settings and paste the generated configuration, while administrators manage company MCP services used by system integrations.

Getting started

The documented path is Docker: download the single-file doca-compose-v0.2.33.yaml release asset into a fixed deployment directory, fill in DOCA_ORIGIN and a DOCA_CREDENTIAL_MASTER_KEY generated with openssl rand -hex 32, then run docker compose pull and docker compose up -d. Administrator creation follows with docker compose exec doca bash docker/admin.sh bootstrap, and a one-click RepoCloud deployment option is also listed.

How it compares

The facts provided name no comparable products, so Doca stands alone in this registry's document management and e-signature category rather than being contrasted with named alternatives. Its distinguishing shape, on the evidence given, is the combination of five document formats in one self-hosted workbench, an MCP service for external AI clients, and a plugin SDK, all under the MIT licence.

When to use it — and when not to

Doca suits people willing to operate it: the Compose file binds 127.0.0.1:39120, so an HTTP(S) reverse proxy is required before a browser can reach it; the master key, project name, database, attachments and the doca_data volume carry the installation across upgrades; and the AI assistant works only after a model and service are configured. Anyone wanting a fully managed service, or expecting email, calendar, membership or content-moderation features out of the box, should look elsewhere, since the default image does not install those plugins and the browser extension is loaded as an unpacked download rather than from an extension store. It is also an early-stage project at version 0.2.x with 504 stars and 4 forks, so teams needing a long operating record should weigh that.

# 1. 在固定部署目录下载单文件配置
mkdir doca && cd doca
curl -fL https://github.com/workhold/doca/releases/download/v0.2.33/doca-compose-v0.2.33.yaml -o compose.yaml

# 2. 用 openssl rand -hex 32 生成主密钥
# 编辑 compose.yaml，填写 DOCA_ORIGIN 和 DOCA_CREDENTIAL_MASTER_KEY
chmod 600 compose.yaml

# 3. 拉取、启动并等待 healthy
docker compose pull
docker compose up -d
docker compose ps

# 4. 初始化管理员，按终端提示输入账号与密码
docker compose exec doca bash docker/admin.sh bootstrap

pnpm install --frozen-lockfile
cp .env.example .env
# 把存储 root 改为本机可写的绝对路径；首次安装替换公开示例主密钥。
pnpm dev