MCP server

Run StoreLab from your agent.

StoreLab's MCP server puts the whole platform in your agent's hands: fourteen tools to draft, publish and read store page tests. You approve what matters; it does the rest.

connect an mcp client
claude mcp add --transport http StoreLab https://mcp.7play.ai/mcp
01setup

However your agent connects.

Claude Code
claude mcp add --transport http StoreLab https://mcp.7play.ai/mcp
Claude desktop and web

Settings, Connectors, add custom connector:

https://mcp.7play.ai/mcp
Codex CLI
[mcp_servers.StoreLab]
url = "https://mcp.7play.ai/mcp"

in ~/.codex/config.toml

Any MCP client

Streamable HTTP at https://mcp.7play.ai/mcp. Sign-in opens in the browser with your StoreLab Google account.

/plugin marketplace add 7play-ai/storelab-plugin then /plugin install storelab@storelab-pluginor the plugin,github.com/7play-ai/storelab-plugin
02the tools

Fourteen tools, every plan sees all of them.

Plans apply per call: upgrade mid-session and the next call works, no reconnect.

search_docsFreeTesting methodology, retrieved into the agent's context.
list_projectsFreeWorkspaces and projects with experiment counts.
list_experimentsFreeStatus, variants and public state per experiment.
get_experimentFreeOne experiment in full: variants, listings, asset refs, public URL.
list_assetsFreeThe project's uploaded creative, ready for reuse.
get_experiment_resultsFreeSessions, install-intent, quality gates and the decision object. Summary by default, full on request.
get_share_linksFreeThe public URL on every plan. UTM channel links need Pro.
get_source_breakdownProInstall-intent split by ad channel.
create_experimentProA brief in, a draft with 2 to 4 variants out.
update_experimentProEdit the draft: scalars or the whole variant set.
upload_assetProReserve a presigned upload, push the file, confirm it.
set_audience_attestationProRecords your general-audience answer. The agent asks; only you decide.
publish_experimentProPreflight first, always. Publishes only after you confirm the exact revision you saw.
record_decisionProWrites the append-only decision. Never changes the live page.
03trust

Built to be given real access.

Publishing always asks first

The agent shows you the readiness preflight and gets your explicit yes before anything ships. A stale preflight is rejected, so nothing goes live that you have not seen.

Your permissions, not more

Tool calls resolve the same user, workspace and role chain as the dashboard. An agent cannot reach a workspace you cannot.

Plans apply per call

Free reads. Pro writes. Upgrade mid-session and the new tools work on the next call.

Least privilege by design

The MCP worker holds no billing keys, no email sending, no publishing signature material. The blast radius of a bad prompt stays small.

Your agent is ready when you are.

StoreLab
store page experiments · lives in your agent over MCP · part of the 7play.ai suite
© 2026 7play.aistore preview pages are used for market research and may redirect to an official app store when available · not affiliated with apple or google