Bosque

Founder

Built Bosque, a workspace for product teams that keeps product decisions and product knowledge in one place. Bosque brings the backlog, roadmap, prioritization, customer feedback, and user research together with a knowledge library and an AI assistant that answers from the team's own documents, with sources. It is source-available under the Elastic License 2.0, so teams can run it on their own infrastructure.

bosque.cc

The Problem

Product knowledge is scattered across too many tools. The backlog lives in one place, interviews in another, customer requests in a shared inbox, and the documents that explain all of it in a drive nobody searches.

The result is that decisions lose their context. A card says what to build, but not why. The customer who asked for it, the interview quote that confirmed it, and the spec that defined it all sit in different tabs. Months later, the same questions get asked again because the reasoning was never kept next to the work.

Cursor for Product Managers

The idea started as a question: what would Cursor look like for product work?

Agentic coding showed that the right environment can make AI genuinely useful. The model sits next to the code, understands the project, and helps you reason inside it. Product managers have no equivalent. They stitch together docs, dashboards, call notes, and intuition, and the AI they use has no idea what their product is.

Bosque is my answer. One working surface for the whole product loop, where every decision points to its evidence and the assistant answers from what the team actually knows.

What Bosque Does

  • Backlog and roadmap. Every card carries its type, priority, effort, owner, dates, and evidence. The same work can be viewed as a ranked table, a phase board, or a timeline.
  • Prioritization and planning. Scoring tables, RICE, and 2x2 boards, with goals and capacity per quarter so the plan matches what the team can deliver.
  • Customer feedback. A shareable feedback portal, a triage inbox, and a link from each request to the card that answers it.
  • Research. Interviews are stored in one place. The moments that matter get highlighted and attached to the cards and feedback they support.
  • Customers. A light record for each customer, trial, prospect, or partner, with everything the team has learned from them.
  • Traceability. For teams working under design controls, needs, requirements, cards, verification, and risks stay connected so the trail is always current.
  • Knowledge Base. Specs, PRDs, guides, and notes live right next to the backlog and roadmap, with comments, versions, and controlled sharing.
  • Knowledge Library and Ask Bosque. Files, websites, YouTube videos, and articles, plus Drive, OneDrive, Notion, and Confluence, become a library the assistant answers from. Every answer shows its sources.

The Product Loop

I organized the product around a simple loop instead of a list of modules:

  1. Plant. Gather feedback, interviews, and documents in one place.
  2. Tend. Score and prioritize together, then set goals and capacity.
  3. Harvest. Run the backlog, write the specs, and share the roadmap.
  4. Grow. Ask questions, read the insights, and keep the knowledge fresh.

Each stage feeds the next. A feedback request becomes a backlog card, a research highlight becomes its evidence, the spec explains the decision, and the assistant can answer questions about all of it later.

Answers Should Cite Their Sources

An assistant that cannot show its work is hard to trust, especially when it is answering questions about policy, customers, or past decisions. Ask Bosque answers only from the Knowledge Library and shows the documents it used next to every reply, so people can check.

Teams stay in control of what the AI can read. Each library item has its own AI toggle, can be scoped to specific departments, and every workspace is private by default.

Source-Available and Self-Hostable

Bosque is source-available under the Elastic License 2.0. Teams can run it for their own organization, including commercially, modify it to fit how they work, and pay no license fee. They bring their own Convex, Clerk, and OpenAI accounts, deploy the backend to Convex, and run the web app from Docker or any Node.js host.

There is also a hosted version with a 14-day free trial for teams that would rather not operate it themselves.

What I Built

  • Turned the "Cursor for product managers" idea into a working product: one workspace for planning, discovery, and knowledge instead of ten tools.
  • Designed the data model around evidence, so backlog cards link to the feedback, research highlights, and requirements behind them.
  • Built the feedback portal, triage inbox, and research highlighting flows that connect customer input directly to the backlog.
  • Built prioritization with scoring tables, RICE, 2x2 boards, and quarterly goals and capacity.
  • Added traceability for regulated teams, connecting needs, requirements, verification, and risks.
  • Built the Knowledge Library and Ask Bosque, with ingestion from files, websites, YouTube, and connected tools, and answers that always cite their sources.
  • Built a collaborative Knowledge Base for specs and PRDs inside the product section.
  • Built it on Next.js, Convex, Clerk, and OpenAI, and packaged it for self-hosting under the Elastic License 2.0.