Agent Connect · Model Context Protocol

Run Opsync with your own AI agent

Connect Claude, Cursor, or any MCP client to Opsync — manage clients, tasks, and reports in plain English.

See how to connectStart free →
What is MCP

The Model Context Protocol is an open, industry-standard way for AI agents to call tools — the same idea as a plugin system, but built specifically for LLMs. Opsync ships an MCP server so your agent can ask Opsync for real data and take real actions, instead of you copy-pasting between a dashboard and a chat window.

What You Can Do

Six tools. One conversation.

🗂️

Client 360°

Pull any client's connected accounts, health score, and stats — one question instead of five dashboard clicks.

✅

Task command

List or create tasks across your team without opening the Kanban board.

📄

Report on demand

Generate and fetch client reports straight from a conversation.

🏢

Agency overview

Agencies, clients, and plan limits at a glance — built for platform admins.

💬

Plain English

No API docs, no auth headers to remember — just ask your agent.

🔒

Fully scoped

Row-Level Security means your agent sees only what your key allows — nothing more.

How To Connect

Three steps. Five minutes.

1

Get your API key

Your agency's Agent Connect key lives in Settings — the same key your agency uses for any internal API access. Copy it once.

2

Pick your agent

Run this once in your terminal:

claude mcp add opsync -- docker run -i \
  -e OPSYNC_AGENCY_API_KEY=<your-agency-key> \
  -e OPSYNC_API_URL=https://opsync.tech \
  opsync-mcp
3

Ask

Try prompts like:

  • “What's BrandMax's health score?”
  • “Create a task: audit PPC for client X”
  • “Pull this week's reports”
Example

What it looks like

Claude Code — connected to opsync-mcp
you ▸ How are my clients doing this week?
agent ▸ Checked agency_clients + agency_profile. 12 active clients, 2 pending tasks overall. One client is below a 90 health score — want me to open a task for it?
you ▸ Yes, and pull their latest report.
agent ▸ Task created. Their most recent report is from this Monday — summarizing it now…
Security
✓Your key stays on your machine — it's never sent anywhere but opsync.tech
✓Client data is RLS-isolated — a client token can only ever see that one client
✓No per-token billing — bring your own AI key, Opsync's price doesn't change
✓SP-API credentials are never exposed to agents — at most a connection status

Agent Connect is included in Growth and Scale

No add-on fee. No per-call metering. It's part of the plan you're already on.

Start free →
FAQ

Questions, answered

Do I need to be technical?+

You need to be able to paste one command into your agent's settings once. After that, it's just conversation — the agent does the rest.

Is my data safe?+

Yes. Requests go straight to the same Opsync API your dashboard uses, over HTTPS, scoped by the same Row-Level Security rules that protect the dashboard and portal today.

Do you charge per AI call?+

No. Opsync never meters AI usage. Your agent uses your own Claude, Cursor, or other model subscription — Agent Connect is included in your Opsync plan at no extra cost.

Can my client use this?+

Yes — client-scoped tools (portal_profile, portal_tasks, portal_reports) work with a client's own portal token, so their agent only ever sees their own account.

Which agents work?+

Any agent or IDE that speaks the Model Context Protocol over stdio — Claude Code, Cursor, and Hermes are documented here, but the server works with any compliant MCP client.