MCP server

Q: Can the assistant change my Shopify data?

No. The API is read-only. The only tool with any effect is trigger_export, which runs an export you configured yourself.

The EcomSolo MCP server lets an AI assistant query your data directly. Connect it once and you can ask, in plain language:

What were my total sales and shipping per store last month?

and get a real answer computed across every Shopify store connected to your account — not a guess, and not one store at a time.

Claude answering a multi-store sales question with EcomSolo data

MCP (the Model Context Protocol) is the open standard assistants use to talk to external systems. Our server is a thin layer over the same /v1 API: identical data, identical permissions, identical rate limits.

Endpoint

https://api.ecomsolo.com/mcp

Authentication is the same API key you use everywhere else — an Authorization: Bearer esk_live_… header. See Authentication for creating one.

The server implements MCP protocol revision 2026-07-28 over Streamable HTTP. It is stateless: there is no session to establish and no handshake to complete.

Connecting

Claude Code (CLI)

One command:

claude mcp add --transport http ecomsolo https://api.ecomsolo.com/mcp \
  --header "Authorization: Bearer esk_live_YOUR_KEY"

Verify with claude mcp list — you should see ecomsolo: ✔ Connected. Then ask your question in any Claude Code session.

claude.ai and Claude Desktop

Custom connectors on claude.ai and Claude Desktop require OAuth, which this server does not offer yet — an API key cannot be entered there directly. Use Claude Code for now; OAuth support is on our roadmap.

Other clients

Any MCP client that supports Streamable HTTP and lets you set an Authorization header works: point it at the endpoint above and pass Bearer esk_live_….

What the assistant can do

Four tools are exposed:

Tool	What it does
`list_datasets`	Lists every queryable dataset and its fields, so the assistant can discover the data model instead of guessing field names.
`run_query`	Runs a SoloQL query across all your stores and returns the rows.
`run_saved_report`	Lists your saved reports, or runs one by id — the reliable way to reproduce a number you already see in the app.
`trigger_export`	Lists your export configurations, or triggers one. Requires a key with the `exports` scope.

Questions that work well

"Compare total sales per store for the last 30 days."
"Which products sold best across all my stores in July?"
"Run my Monthly Revenue report and summarise what changed."
"How much did I collect in shipping last quarter, by country?"

The assistant will usually call list_datasets first to learn the field names, then run_query. That is expected, and it is why answers are grounded in your real data rather than invented.

What it cannot do

The API is read-only. There are no tools to create, update or delete anything in your stores. An assistant connected to EcomSolo can read your data and trigger an export you have already configured — nothing else.

Exports are asynchronous: trigger_export returns a job id as soon as the job is queued, not when the file is ready. Poll with GET /v1/export-jobs/{id} to download it.

Permissions, scopes and limits

Everything that governs the REST API governs MCP identically, because it is the same code underneath:

Plan — the API is included on Advanced and above.
Scopes — read_data for queries and reports, exports for triggering exports. A key missing a scope gets a clear message the assistant can relay.
Store restrictions — if a key is limited to certain stores, MCP sees only those stores. The assistant cannot widen that.
Rate limits — MCP calls count against the same per-minute budget as regular API calls. See Rate limits.
Revocation — revoke the key and the connection stops working immediately.

Security

Treat an MCP connection as what it is: you are granting an AI assistant read access to your business data. A few things worth knowing.

Create a dedicated key for the assistant rather than reusing one that is already wired into a production integration. You can then revoke it without breaking anything else.
Grant the narrowest scopes that do the job. If you never want the assistant triggering exports, do not give it the exports scope.
If the key should only see some stores, restrict it to those stores when you create it.
The conversation, including the data returned, goes to whichever AI provider you connected. That is inherent to the arrangement, not specific to EcomSolo — but it is worth a deliberate decision rather than a default one.

FAQ

Does this cost extra? No. MCP is part of the API, included on Advanced and above at no additional charge. There are no per-call fees.

Do I need a separate key for MCP? Technically no — any API key works. In practice yes, we recommend one, so it can be revoked independently.

Can the assistant change my Shopify data? No. The API is read-only. The only tool with any effect is trigger_export, which runs an export you configured yourself.

Which stores does it see? Every store connected to your EcomSolo account, unless the key is restricted to a subset. A single question is answered across all of them at once.

What if the assistant gets a field name wrong? It gets a validation error explaining the problem and can correct itself — usually by calling list_datasets to find the right name.