Feature request for product/service
– Other –
Describe the request
Recently, Rowboat was released. It is positioned as an assistant agent that, in theory, could replace a personal secretary. I will not go into its general capabilities here — they are described in detail in the repository:
What I find particularly interesting is its architecture. The agent accumulates knowledge and builds a structured note tree in Obsidian, where notes can reference and extend one another. This enables, at least in theory, a persistent analytical layer over a large codebase: created once, then incrementally updated as the project evolves.
Such an approach could eliminate the need to repeatedly reconstruct global context from scratch. It also reduces reliance on ad-hoc, quickly outdated .md reports. With a maintained knowledge graph, higher-level tasks could be delegated without assembling full context manually each time or worrying that the model will overlook subtle constraints.
An important aspect is that the generated knowledge base remains transparent: it is fully readable and editable by humans, and suitable for cross-verification and manual refinement.
In the context of Cursor, this model seems especially promising. Cursor already provides BugBot, a background agent that analyzes the codebase for bugs. A complementary agent dedicated to building and maintaining a structured knowledge base could significantly enhance long-term project understanding.
Additionally, within an agent-based workflow, cheaper models could be used to retrieve and assemble context from the note graph, delegating higher-level reasoning to more capable models. This would resemble Cursor’s Plan Mode, which has already demonstrated strong effectiveness.
Such an architecture would shift AI-assisted development from stateless prompting toward persistent, evolving project memory.
