Why Agency Operators Use Muse with HighLevel Agency & Builder MCP
Muse excels at orchestrating multi-step operator workflows, but needs deterministic dry-run plans, structured JSON output, and post-write read-back verification so autonomous runs never silently corrupt client sub-accounts.
Deterministic Plan / Apply Execution for Muse.app
- Ships with a native `.muse-plugin/plugin.json` manifest and skill definition (`docs/publishing/generated/muse/`).
- Every mutating tool defaults to `dryRun: true` and returns a structured JSON plan before applying changes with `yes: true`.
- Supports both Hosted OAuth 2.1 (`https://ghlagencymcp.com/mcp`) and local `ghl-agency` binary execution.
- Automatically redacts PITs, JWTs, and sensitive credentials from tool outputs.
Side-by-Side Technical & Pricing Comparison
| Capability / Requirement | Unverified Script Wrappers | Browser Automation in Muse | HighLevel Agency & Builder MCP |
|---|---|---|---|
| Execution Speed & Reliability | ⚠️ Node/Python runtime drift | ❌ Brittle DOM selectors & login prompts | ✅ Sub-millisecond compiled Rust binary + HTTP MCP |
| Dry-Run Plan + Read-Back Contract | ❌ Trusts raw HTTP 200 | ❌ Visual guessing | ✅ Cryptographic state hashing & live read-back |
CLI & MCP Configuration
# Generate Muse.app MCP configuration
ghl-agency init-config --client muse
# Or add Hosted Streamable HTTP MCP in Muse Settings:
https://ghlagencymcp.com/mcpCopy-Paste AI Prompt
Inspect all sub-accounts in our agency context, run a snapshot diff against our Master Snapshot, and generate a dry-run remediation plan for any location with missing custom values.
Frequently Asked Questions
Does the Muse plugin include operator skill instructions?
Yes. `docs/publishing/generated/muse/skills/highlevel-agency-mcp/SKILL.md` provides Muse with exact tool schemas and dry-run safety rules.
Can Muse run both local CLI and remote HTTP MCP?
Yes. `ghl-agency init-config --client muse` supports both local `stdio` and `--http` modes.