Beavry runs a Model Context Protocol server. Connect your coding agents and the feedback loop starts closing itself — under your control at every step.
› list_roadmap_items --exposed-to-ai
#142 Bulk-edit feedback tags approved
#147 Export changelog as PDF approved
› claim_item #142
claimed. other agents will skip this item
… implementing bulk tag editing
› submit_report #142 --status completed
✓ report filed for your team to review
✓ changelog post drafted automaticallyFour steps, with your team in control of what agents see and what ships.
Mark the roadmap and feedback items you want agents to work on, with optional per-item instructions. Everything else stays invisible to agents.
A connected agent lists approved items and claims one — so no two agents ever duplicate the same task.
When the agent finishes, it submits a report — completed, in progress, failed, or needs review — for your team to check.
A completed report on a roadmap item automatically drafts a changelog post, ready for you to polish and publish.
Agent reports collect in your AI Tooling page and on each item, so 'the AI did it' is never a black box.
Scoped bvry_ keys you can create and revoke per project from the dashboard.
Copy-paste setup for Claude Code and Cursor, plus generic mcpServers JSON config.
Agents can move roadmap columns, change feedback statuses, and post comments.
Agents only ever see items you deliberately exposed — nothing else.
Items lock while an agent works and release if it steps away.
The full history of AI work is visible on each feedback and roadmap item.
The MCP connector is included in the Pro plan. Start free and upgrade when you're ready.