LLM-maintained Project Knowledge Bases
TrialFactorial rates LLM-maintained knowledge bases as Trial (2026). Each of our client workspaces now carries such a knowledge base, set up from an internal template with skills for ingesting, authoring, linting and querying.
A project knowledge base kept as Markdown in git, structured so that an AI agent can ingest new material (meeting notes, tickets, documents), write and link notes, check them for consistency and answer questions from them. People review the changes like code.
Updates
Each of our client workspaces now carries such a knowledge base, set up from an internal template with skills for ingesting, authoring, linting and querying. It gives every agent session, and every new team member, the same project context, and replaced our earlier per-person onboarding for AI tools. We are still learning how much curation it needs to stay trustworthy.