Kanban Agent Long Term Memory. Visualized

What is Kanban Agent Long Term Memory. Visualized?

CWIM Kanban gives your AI agent a persistent memory layer. It remembers what you were working on, recalls context automatically, and shows everything on a live dashboard. It works with Claude Code, OpenCode, and any Git repository.

How to use Kanban Agent Long Term Memory. Visualized?

Install globally with npm install -g @cwim/kanban then run kanban to launch the dashboard. For AI integration, add the MCP configuration to your Claude Code or OpenCode settings and place a CLAUDE.md or AGENTS.md file in your project root with the provided workflow instructions.

Key features of Kanban Agent Long Term Memory. Visualized

• Smart context recall via task_recall
• Timestamped note appending without overwriting
• Per‑project session isolation with Git auto‑detection
• Real‑time visual dashboard (4 columns, tag support)
• Local‑first JSON storage at ~/.kanban/sessions/
• Full suite of MCP tools for task and session management

Use cases of Kanban Agent Long Term Memory. Visualized

• Resume a complex refactoring task exactly where you left off
• Track multiple project tasks in parallel without context bleeding
• Allow AI agents to autonomously log discoveries and blockers
• Visualize and filter work progress across all your repositories

FAQ from Kanban Agent Long Term Memory. Visualized

Where is my data stored?

All data is stored locally in ~/.kanban/sessions/ as human‑readable JSON files. No cloud accounts or network connections are required.

What are the system requirements?

Node.js 18+ is required. Git is optional (used for automatic repository detection). Claude Code or OpenCode are optional – the dashboard works independently.

Does the dashboard require an internet connection?

No. The MCP server and HTTP server run locally and communicate via stdio and file polling. No external services are used.

How does session isolation work?

Each Git repository or project directory gets its own session. The tool auto‑detects the active repository when you run commands inside it, preventing context from leaking between projects.

What MCP tools are available?

The server provides task_recall, task_create, task_append_note, task_update, task_move, task_delete, task_list, task_get, session_list, and session_switch.