OpenClaw plugin that connects AI agents to Kanvas — your company's nervous system. Kanvas is the operational engine that connects all your data, tools, and workflows. For AI agents, it's what lets them actually run your business — not just talk about it.

This plugin gives agents direct access to CRM, inventory, orders, and messaging — 53 tools with auto-login, built-in system prompt context, and domain-specific skills.

openclaw plugins install @kanvas/openclaw-plugin

openclaw plugins update kanvas

In your OpenClaw config file:

{
  plugins: {
    entries: {
      "kanvas": {
        enabled: true,
        config: {
          xKanvasApp: "your-app-id",
          email: "[email protected]",
          password: "agent-password",
          // Optional:
          // xKanvasKey: "your-app-key"  // needed for kanvas_send_anonymous_email
          // xKanvasLocation: "branch-uuid"
        }
      }
    }
  }
}

The API URL is preconfigured. The plugin authenticates automatically on the first tool call using the Kanvas login mutation.

The plugin ships with domain-specific skills — operational playbooks that teach the agent best practices for each use case. Skills are auto-loaded when the kanvas plugin is configured.

Skills live in skills/<skill-name>/SKILL.md and follow the OpenClaw skills format. They inject operational guidance into the agent's context automatically.

The agent-setup/ directory contains step-by-step runbooks for deploying specific agent types. These are human-facing documentation — not auto-loaded by the agent.

Sales / Deal Architect Autonomous lead generation, cold outreach, pipeline management, follow-up scheduling, and daily reporting. See agent-setup/sales.md for the full setup protocol.

Operations Agent Inventory monitoring, order tracking, stock alerts, and cross-domain reporting.

Customer Success Agent Post-sale follow-ups, onboarding workflows, satisfaction tracking via CRM notes and events.

Data Entry / Admin Agent Bulk lead import, contact enrichment, file attachments, and CRM hygiene.

Each agent type uses the same plugin and tools — the difference is the system prompt, skills, and heartbeat configuration.

The plugin injects tool documentation automatically. But the agent's base system prompt (configured in OpenClaw) should define its role. Example for a sales agent:

You are a sales and operations agent for [Company Name]. You manage leads,
inventory, and orders using the CRM.

Your responsibilities:
- Register and manage leads in the CRM pipeline
- Send outreach emails and follow up with prospects
- Log all activity as CRM notes and messages
- Schedule follow-ups as calendar events (never in local memory)
- Send daily pipeline summaries to the team

When users ask you to do any of these things, use the kanvas_* tools to act
on it directly. Don't just describe what you would do — actually do it.

Messages act as NoSQL-like document storage — the message field accepts any JSON structure.

The plugin supports three auth modes:

With email/password, the plugin calls the Kanvas login mutation on the first tool invocation and caches the bearer token — no hardcoded tokens needed.

*Required when using email/password auth (recommended). Not needed if using bearerToken or app-key mode.

All fields also fall back to KANVAS_* environment variables.

npm run build     # Compile TypeScript to dist/
npm run check     # Type-check without emitting
npm run dev       # Watch mode

Test against the real API:

KANVAS_X_APP=your-app-id \
[email protected] \
KANVAS_PASSWORD=yourpassword \
  npx ts-node --esm scripts/smoke-test.ts

npm version patch   # bumps 0.1.x → 0.1.x+1 and creates a git commit + tag
git push && git push --tags

GitHub Actions builds and publishes to npm automatically on any v* tag.

Then on the agent machine:

openclaw plugins update kanvas

MIT