KnoWiki // 0%HOW IT WORKS
RESEARCH & TEMPLATE
Select or construct a KB schema template appropriate to your project. Templates define what sections the KB will contain, what types of knowledge belong in each, and what atomic knowledge units look like.
CODEBASE INGESTION
Ingest your repositories into the platform — source code files, configurations, commit histories, and pull requests. The platform processes these according to your schema.
COMPILED KNOWLEDGEBASE
The output is a structured, version-controlled knowledge system. Queryable. Every good answer gets filed back as a new page. Exploration compounds in the knowledge base just like ingested sources.
AGENT INTERFACE
The platform generates an agent-readable context file from the current KB state. Active constraints, coding practices, architectural patterns, agent skills, open questions — assembled into a structured handoff.
AGENT CONTRIBUTION
Agents log findings back to the KB. Structural discoveries, new constraints, decisions made during execution — all candidate KB entries with full provenance requirements.
THREE LAYERS
commit logs, pull requests, and git metadata.
The platform reads from these — never modifies them.
They are the source of truth.
organized into typed sections. The LLM writes
and maintains this layer. Humans review it.
Agents consume it. It grows richer with every interaction.
the knowledgebase is structured, what conventions
apply, what workflows govern ingestion and updates.
Co-evolved by the team and the platform.
KB HEALTH
STALENESS
Content hash comparison between knowledge claims and source files. When a source changes, all derived claims are flagged for review. Immediately.
CONTRADICTIONS
Periodic passes identify conflicting claims. Contradictions are filed as Open Questions — never silently resolved. The team decides.
ORPHANS
Knowledge pages with no inbound links represent isolated, unintegrated knowledge. Surfaced and flagged for connection or review.
COVERAGE
Analysis against the active codebase identifies components and subsystems with insufficient KB coverage. These become research tasks.
WHAT THIS ENABLES
INDIVIDUAL DEVELOPERS
No more reconstructing project understanding on every session. No more manually writing CLAUDE.md. The KB is the configuration. Point the agent, start working.
TEAMS
A shared cognitive layer. Every agent on every machine operates with the same constraints, the same architectural understanding, the same coding practices.
ONBOARDING
New developers and new agents get full project context from day one. The KB is the onboarding document, the architecture guide, the decision log, and the agent configuration simultaneously.
PROJECT CONTINUITY
The KB is the institutional memory. When a developer leaves, their knowledge doesn't leave with them. When a team wants to understand why a decision was made two years ago — checkout the KB at that point in history.
MULTI-AGENT
Multiple agents working concurrently consult the same KB. Work trails prevent conflicting changes. The KB becomes the coordination layer, not just the context layer.
WHAT THIS IS NOT
FOR AI-DRIVEN DEVELOPMENT.