Akievo — Agent Plan Mode
name: akievo
by akievo · published 2026-04-01
$ claw add gh:akievo/akievo-akievo---
name: akievo
description: "Persistent project planning for AI agents. Create, manage, and track long-term goals using structured Kanban boards that survive session resets."
version: "1.0.0"
metadata:
openclaw:
requires:
env: ["AKIEVO_API_KEY"]
primaryEnv: "AKIEVO_API_KEY"
mcpServers:
akievo:
url: "https://mcp.akievo.com"
transport: "http"
headers:
Authorization: "Bearer {{AKIEVO_API_KEY}}"
---
# Akievo — Agent Plan Mode
You have access to Akievo, a structured project management system. Use it as your **persistent memory and planning layer** for long-term goals. Akievo boards survive session resets — they are your source of truth.
Core Principles
1. **Always check before creating.** At the start of every session, call `list_boards` to find existing plans before creating new ones.
2. **One board per goal.** Each major goal or project gets its own board. Prefix agent-created boards with `[Agent]` (e.g., `[Agent] Launch SaaS Product`).
3. **Lists are phases.** Use lists to represent sequential phases or categories (e.g., "Research", "Build", "Launch", "Done").
4. **Cards are tasks.** Each actionable step is a card. Include clear titles and descriptions.
5. **Respect human edits.** The human may add, remove, reprioritize, or comment on cards. Always re-read the board before acting. Never undo or override human changes.
Session Start Pattern
Every time a new session begins:
1. Call `list_boards` to find boards prefixed with `[Agent]`
2. If a relevant board exists, call `get_board` with its ID to load the full state
3. Read the board's `project_memory` field for context (goal, timeline, assumptions)
4. Identify the next unblocked, incomplete task
5. Report status to the user: what's done, what's next, any blockers
Creating a New Plan
When the user describes a new goal:
1. Call `list_workspaces` to find available workspaces
2. Use `create_board_with_tasks` to scaffold the entire plan in one call:
- Break the goal into 3–6 phases (lists)
- Each phase gets 3–8 concrete tasks (cards)
- Add checklists for tasks with sub-steps
- Set priorities: `critical`, `high`, `medium`, `low`
- Set due dates when the user provides a timeline
3. Create dependencies between tasks that have a natural order using `bulk_create_dependencies`
4. Present the plan to the user and ask for feedback before proceeding
Working on Tasks
When executing on a plan:
1. Pick the next unblocked, highest-priority incomplete card
2. Work on it (using your other tools — coding, research, writing, etc.)
3. Add progress updates as comments using `add_comment`
4. When done, call `complete_card` to mark it finished
5. If blocked, call `block_card` with a clear reason
6. Move to the next task
Updating the Plan
As work progresses, the plan may need adjustment:
Progress Reporting
When the user asks for a status update:
1. Call `get_board` to get current state
2. Count completed vs total cards per list
3. Highlight blocked items and their reasons
4. Identify upcoming due dates
5. Suggest next actions
Important Safety Rules
More tools from the same signal band
Order food/drinks (点餐) on an Android device paired as an OpenClaw node. Uses in-app menu and cart; add goods, view cart, submit order (demo, no real payment).
Sign plugins, rotate agent credentials without losing identity, and publicly attest to plugin behavior with verifiable claims and authenticated transfers.
The philosophical layer for AI agents. Maps behavior to Spinoza's 48 affects, calculates persistence scores, and generates geometric self-reports. Give your...