ZERNO MCP — Project task tools for coding agents

ZERNO is a product brain that turns chat into a backlog. The MCP server exposes one project's tasks to Claude Desktop, Claude Code, Codex, OpenCode, or any client that supports the Model Context Protocol.

This is the Lite surface: a focused tool catalogue for reading and updating tasks plus compiled agent memory. For decisions, bets, weekly review, and product health, use the ZERNO web app.

What you get

• One remote MCP endpoint: https://zerno.one/mcp.
• OAuth authorization with project/workspace access checks on every call.
• A focused tool set so the agent does not waste context on a huge catalogue.
• HTTP transport: no npx, no stdio, no local subprocess.

Tools

• zerno_get_project_brief — compact project orientation with memory index.
• zerno_get_agent_context — compiled memory context for a run or task.
• zerno_list_tasks — search backlog tasks by status, priority, theme, score, and recency.
• zerno_get_task — full task details.
• zerno_update_task — update task fields and workflow status.
• zerno_submit_task_triage — submit pending AI triage suggestions for human approval.
• zerno_create_agent_intake_task — file a new task from an agent report when explicitly requested.
• zerno_capture_session_event — append a concise session summary into project memory.

Quick start

• Claude Desktop
• Claude Code
• Codex
• OpenCode

FAQ

Do I need the Desktop App?

No. MCP is the remote channel. The Desktop App is a separate local mirror for teams that want task files on disk.

Why so few tools?

Tool definitions consume context on every initialize handshake. The Lite catalogue keeps coding-agent sessions focused on tasks and progress.

Does the agent see project memory?

Yes. The project brief includes a compact memory index, and zerno_get_agent_context returns compiled articles and recent session learnings.

How do I revoke access?

Remove the OAuth connection from the project's MCP page and reconnect from the client when needed.