Build your AI workforce
Docana is the home for your AI workforce. Create an employee for one task, automate a complex process, or connect specialists into a team. Build in Docana, work from an AI client through MCP, or manage workflows from your own code. Their knowledge, configuration, and access stay together in Docana.
These docs use agent for the configurable workflow behind an employee. Start with the task you want to automate, then connect the documents and tools it needs.
Describe a job, add knowledge, test expectations, and work together.
Run agents, automate document uploads, and add tests to your workflow.
Find authentication, request examples, response schemas, and errors.
Connect an MCP client to search knowledge and manage agents.
How Docana fits together
Your knowledge supplies the context. An application controls which knowledge and agents its users can access. People work with those employees through channels, connected AI clients, or the API.
Start with Core Concepts to understand libraries, collections, applications, agents, templates, and channels.
Choose what to build
| Your goal | Start here |
|---|---|
| Create an employee or automate a team workflow | Employees |
| Answer questions from company documents | Quickstart |
| Keep documents in sync with their source | Connectors |
| Review the same fields across many documents | Document Insights |
| Query rows and calculate totals from a CSV | Datasets |
| Build a workflow with explicit steps and tools | Creating an Agent |
| Generate documents in a repeatable format | Creating a Template |
| Create and prepare an employee through REST, MCP, or CLI | Employees from Code |
| Keep workflows and tests under version control | Agent Projects |
| Check agent changes before publishing | Test Agents in CI |
| Run an agent on a schedule or document event | Routines |
| Call an agent from your backend | Run Agents from Your Systems |
Build, verify, then share
Start with a small set of representative documents and questions. Check the sources in the answers and add test cases before you share an agent. Tests help you catch regressions when instructions, tools, or documents change. They do not guarantee that every future answer will be correct.
Publish the reviewed agent before starting a working chat or calling it from your system. An employee draft shows Review & publish in the team list and Publish for chat in preparation. Once published, use it from your AI client via MCP, or connect a web widget, WhatsApp, Slack, or Microsoft Teams. Use Insights to review conversations and identify what to improve.
Connect the sources the work needs
A preparation brief describes the assignment. Collections and connectors supply lasting knowledge. Choose managed connectors for supported sources, document ingestion for files arriving from your own systems, website access for allowed sites, or MCP servers for external provider tools.
Your code can search an application's knowledge or read a document's extracted content. Both respect the caller's access. See Core Concepts for the relationship between applications, libraries, and collections.
Work from code
Choose the interface that fits your workflow:
- REST API: integrate Docana into your backend with explicit HTTP requests.
- CLI: manage agents and documents from a terminal, including version control and CI.
- Docana MCP: give an AI client tools to work in your Docana workspace.
- MCP Servers: connect external tools that your Docana agents can call.
Set up your workspace
Administrators can add users and organize teams, assign access to libraries and applications, and create scoped API keys. Start at Applications or Knowledge Libraries in the platform.
For help with an existing workspace, contact support. To discuss a new deployment, request a demo.