Skip to content

Agent Integration

Start with Connect OpenClaw or Cursor for step-by-step setup, read-only verification, copy-paste AI prompts, and optional Cursor attachment MCP configuration. These guides are also bundled inside the app under DOCS → Connect.

HoneyDew is designed to be used by AI agents as much as by humans. There are three ways to connect your agent, depending on your setup.

Prerequisites

  • HoneyDew must be running (./start.sh).
  • No API key is required — the API is open on localhost.
  • If your agent runs on a different host, set SMARTIFY_API_URL to the correct base URL.

OpenClaw Skill

Best for: OpenClaw users who want fully automatic agent integration.

Install the HoneyDew skill from ClawHub:

clawhub install honeydew

Once installed, your agent knows how to manage your board. Ask it anything:

  • "Create a high-priority task: 'Deploy v2'."
  • "Show me all overdue cards."
  • "Move the 'Fix login bug' card to In Progress."
  • "Transfer everything in Done back to me for review."
  • "What's on the board right now?"

The agent calls the HoneyDew API automatically — no manual configuration needed beyond the install.

See the skill README for details on capabilities and troubleshooting.

Share the Skill Context

Best for: Any agent (Claude, GPT, Cursor, Copilot, etc.) — just give it the context file.

The file skills/honeydew/SKILL.md contains everything an agent needs: when to use it, how to connect, the full API reference, priority values, and curl examples. Share it with your agent by:

  • Dropping it into your system prompt as a context block.
  • Attaching it as a file in a conversation.
  • Copying the relevant sections into your agent's instructions.

Your agent will know how to call the API, create cards, move them between columns, transfer ownership, and more — with no additional setup.

What's in SKILL.md?

The file covers:

  • When to use — Creating, updating, moving, or deleting cards; managing boards and labels; checking overdue or urgent items.
  • Connection — Base URL (http://localhost:8000), env override (SMARTIFY_API_URL), health check (GET /health).
  • Full API table — Every endpoint with method, path, and required fields.
  • Priority values — 1 = Low, 2 = Medium, 3 = High, 4 = Urgent.
  • Curl examples — Create a card, move it, transfer it.
  • Python tools — Optional higher-level interface via kanban_tools.py.

Python Tools Module

Best for: Scripted integrations or agents that run Python directly.

The tools/kanban_tools.py module provides a higher-level interface on top of the REST API:

from kanban_tools import KanbanTools, Priority

kanban = KanbanTools()

# Agent creates a task for itself
card = kanban.create_task(title="Refactor auth module", priority=Priority.HIGH)

# Work on it
kanban.mark_in_progress(card["id"])

# Hand it back to the human for review
kanban.assign_to_user(card["id"])

# Human approves — mark done
kanban.mark_done(card["id"])

The module reads profile IDs and board config from config.json automatically. Override the API base URL with SMARTIFY_API_URL if the server isn't at http://localhost:8000.

For the full method reference, see Python Tools.

Which Option Should I Use?

Scenario Recommendation
You use OpenClaw Install the skill — one command, zero config.
You use a different AI agent (Claude, GPT, Cursor, etc.) Share SKILL.md as context.
You want programmatic / scripted control Python tools module.
You want raw HTTP control Use the REST API directly.

File attachments and storage

See Attachments and S3-compatible storage for issue/Epic uploads, managed-object APIs, native MCP, host file roots, limits, retention and backup/restore.