Knowledge Base
AI agents are only as good as the context you give them. The knowledge base is where you store architecture decisions, API specs, and domain knowledge that agents reference when implementing. Without this documented context, agents guess. With it, they make informed decisions based on your reality.
Almirant's knowledge base is an integrated documentation system where your team can create, organize, and search documents in Markdown format. Documents are grouped into hierarchical categories, linked to projects and work items, and serve as additional context for the AI.
What is a document
A document is a content page written in Markdown. It can be a technical specification, a process guide, meeting notes, architecture decisions, or any knowledge your team needs to share and reference.
Each document belongs to a category and can be linked to one or more projects.
Creating a document
Click New document from the documents section. Fill in the fields:
| Field | Required | Description | Example |
|---|---|---|---|
| Title | Yes | Descriptive name for the document | "Production deployment guide" |
| Content | No | Document body in Markdown | Formatted text, lists, tables, etc. |
| Category | No | Category within the hierarchical tree | "Technical guides > Infrastructure" |
| Project | No | Project the document is linked to | "Mobile App v2" |
Markdown editor
The editor supports full Markdown syntax:
- Headings (
#,##,###) - Lists -- ordered and unordered
- Tables
- Code blocks with syntax highlighting
- Bold, italic, and other inline formatting
- Links and images
A word counter is displayed at the bottom of the editor to give you a reference for the document's size.
Hierarchical categories
Documents are organized in a category tree with unlimited depth. This lets you create structures as detailed as you need:
Documentation
├── Technical guides
│ ├── Infrastructure
│ ├── Database
│ └── APIs
├── Processes
│ ├── Onboarding
│ └── Release management
├── Architecture decisions
└── Meeting notes
├── Sprint reviews
└── Planning
Creating a category
- In the documents sidebar tree, click New category
- Enter the category name
- Optionally, select a parent category to create subcategories
Tree navigation
The left sidebar of the documents section shows the full category tree. Click on a category to see the documents it contains. Categories with subcategories can be expanded and collapsed.
Full-text search
The document search lets you find content by full text. It searches both the title and the body of the document, returning results sorted by relevance.
Type your query in the search bar of the documents section to filter the list in real time.
Linking with work items
Documents can be bidirectionally linked with work items. This is useful for:
- Associating a technical specification with the story that implements it
- Linking an architecture decision to the epic that originated it
- Connecting meeting notes with the tasks that arose from them
Linking from the document
In the document detail view, find the Linked work items section and add the work items you want to associate.
Linking from the work item
From a work item's detail view, you can link existing documents in the corresponding section.
The linking is bidirectional: if you link a document to a work item, the work item will also show the reference to the document.
Pinning
You can pin important documents so they appear highlighted at the top of the list. This is useful for frequently referenced documents such as style guides, critical processes, or onboarding documents.
To pin a document, use the pin icon in the list or in the document detail view.
Cross-project grouping
Documents can be linked to multiple projects, allowing you to create cross-cutting documentation that applies to several teams or initiatives. For example, a "Code standards" document can be linked to all development projects.
Version history
Almirant automatically saves previous versions of each document. Every time you save changes, a new version is stored in S3. You can:
- View a document's version history
- View the content of previous versions
- Identify when changes were made and what was changed
Documents as context for AI
Knowledge base documents can serve as additional context for AI agents connected via MCP. When an agent needs to understand a specification, a process, or an architecture decision, it can query the relevant project documents.
This is especially powerful when you combine documents with work items: the AI can read the technical specification linked to a work item before implementing it.
Documents are available via MCP:
| Tool | Description |
|---|---|
list_documents | Lists documents with filters by project and category |
list_document_categories | Lists the available hierarchical categories |
Documents accessible via MCP allow an AI agent to query the team's knowledge base to make more informed decisions during implementation. For example, before implementing a work item, the agent can search for relevant technical specifications or architecture decisions.
Related: Projects | Work Items