Knowledge base
The Knowledge Base is where you maintain reference documentation about your project’s architecture and systems. Unlike task-level specs (which describe a single piece of work), concepts document cross-cutting topics that span many tasks and evolve independently.
When to use concepts
Use concepts for topics that multiple tasks touch but no single task owns:
- System architecture — “Checkout Pipeline”, “Auth System”, “Notification Service”
- Domain models — “User Permissions Model”, “Billing Domain”
- Integration guides — “Jira Sync Architecture”, “Calendar Integration”
- Runbooks — “Deploy Process”, “Incident Response”
Creating a concept
Open the Knowledge Base tab in your project. Type a name in the “Add concept…” field at the bottom and press Enter.
Each concept has:
- Title — double-click to rename inline
- Icon and color — click the avatar to pick an icon and palette color
- Description — a brief summary of scope and purpose
- Overview — what this concept is, its boundaries, constraints, and context (markdown with Mermaid diagrams)
- Architecture — how it works: structure, data flow, implementation details, diagrams
Referencing concepts from tasks
You can link to a concept from any task’s spec or design doc using inline references:
See {{concept:"Checkout Pipeline"}} for the full architecture.This renders as a clickable card showing the concept’s icon, title, and an excerpt of the overview. Clicking it navigates to the Knowledge Base tab.
The concept reference picker is also available in the markdown editor toolbar — click the book icon to browse and insert a reference.
:::note Concept references inside a concept’s own Overview or Architecture sections are disabled to prevent circular references. :::
Organizing concepts
- Search — filter concepts by title or description using the search bar
- Expand / Collapse all — toggle all concept cards at once
- Archive — soft-delete a concept while preserving its content. Archived concepts are hidden by default but can be restored.
- Delete — permanently remove a concept and all its content (overview, architecture, diagrams, decisions)
Both archive and delete show a confirmation dialog.
Using concepts with Claude Code
Claude Code can create and manage concepts through the MCP tools:
"Create a concept called 'Checkout Pipeline' and document
how the checkout flow works."Claude Code calls plenar_create_concept to create the concept and write the overview and architecture documentation in a single step. It can also list existing concepts with plenar_list_concepts, read one with plenar_get_concept, revise content with plenar_update_concept, and shelve stale ones with plenar_archive_concept. The same tools let an agent insert concept references into task specs as it works.
Common questions
When should I write a concept instead of a task spec? Write a concept when the topic spans many tasks and no single task owns it — architecture, domain models, integration guides, runbooks. Keep problem-and-solution detail for a single piece of work in that task’s spec and design doc.
What’s the difference between archiving and deleting a concept? Archiving soft-deletes the concept while preserving all its content — you can restore it later. Deleting permanently removes the concept and everything in it (overview, architecture, diagrams, decisions). Both show a confirmation dialog first.
Can I link a concept from a task? Yes. Use the inline reference {{concept:"Title"}} in any task’s spec or design doc, or insert one from the markdown editor toolbar. It renders as a clickable card.
What to do next
- Write a project poster — capture the one-page problem and approach that frames the concepts in your Knowledge Base.
- Goals & milestones — structure the outcomes that your concepts describe how to build.
- Track progress with Claude Code — the daily agent workflow that keeps specs, design docs, and concepts current as you work.