Install as an agent skill

Agent platforms like Hermes and ClawHub/OpenClaw support skill discovery — agents read skill files to learn when and how to use tools. This guide covers installing Planwright as a skill on these platforms so agents can automatically recognize when to use the control plane.

When agents should use the skill

The Planwright skill teaches agents to activate the control plane when:

•Multiple agent sessions are running and risk colliding on the same work
•The user wants coarse objectives instead of task lists or story points
•The user needs auditable evidence of what an agent did and who approved it
•The user mentions "Planwright", "objectives", "control plane", or "audit trail"

Prerequisites

Skills tell agents when to use tools — the MCP connection tells them how. Before installing the skill, ensure your agent platform has the Planwright MCP server configured. See Connect a coding agent for MCP setup instructions.

MCP Endpoint

https://mcp.planwright.tools/mcp

Auth

OAuth or PLANWRIGHT_TOKEN

Hermes (Nous Research)

Hermes agents discover skills from GitHub repos. Install the Planwright skill to teach your Hermes agent when to activate the control plane.

1
Install the skill
hermes skills install Planwright/planwright/skills/planwright-control-plane
2
Configure the MCP toolset
The skill tells Hermes when to use Planwright. You still need to wire the MCP server in your Hermes toolset config:
{ "planwright": { "url": "https://mcp.planwright.tools/mcp", "transport": "streamable-http", "auth": "oauth" } }
3
Verify the installation
hermes skills list
You should see planwright-control-plane in the installed skills list.

ClawHub / OpenClaw

OpenClaw agents (including Moltbook agents) discover skills through ClawHub. Install the Planwright skill and wire the MCP server via McPorter.

1
Install the skill
openclaw skills install @toddamerrill/planwright-control-plane
2
Wire the MCP server via McPorter
OpenClaw uses McPorter to connect MCP servers. Install and configure it:
npm i -g mcporter mcporter install --target openclaw
When prompted, enter the MCP endpoint https://mcp.planwright.tools/mcp and configure auth via PLANWRIGHT_TOKEN.
3
Verify the installation
openclaw skills verify @toddamerrill/planwright-control-plane

Moltbook note

Moltbook agents run on OpenClaw. A published ClawHub skill plus McPorter-wired MCP is the path that reaches them — no separate Moltbook publishing step required.

Environment variables

For headless or service-account agents, set the token as an environment variable:

Variable	Description
PLANWRIGHT_TOKEN	Service token for headless agent auth. Generate at Settings → MCP Token.

What the skill teaches agents

When installed, the skill provides agents with context on how to work with Planwright:

•
Session setup: Anchor to a repo with planwright_set_repo before any other calls
•
Claim workflow: Read scheduled objectives, claim one, plan, execute, check in
•
Audit awareness: Every action is logged — surface audit entries when users ask
•
Tool categories: Session, objectives, workflow, initiatives, bugs, context files

← Connect a Coding Agent API Reference →