Agents and MCP · LiquidUser

Connect with a workspace key

In Settings, open Agents and create a key. Choose what it can do and when it expires, then add it to your MCP client. Copy the key when you create it: LiquidUser shows it only once.

Endpoint: https://mcp.liquiduser.com/mcp
Transport: Streamable HTTP
Header: Authorization: Bearer lu_sk_…

Your client needs to support Streamable HTTP and a custom Authorization header. Where you enter these details differs from client to client.

What an agent can do

A key gets only the scopes you choose. Each tool on the server needs one scope, listed next to it here.

Scope	Allows	Tools
`market:read`	See what LiquidUser found for this workspace, with the source for each item.	`search_people` `search_problems` `search_conversations` `search_companies` `get_person` `get_problem` `get_company` `get_opportunity` `get_evidence` `get_daily_brief`
`memory:read`	See what LiquidUser knows about your product and audience.	`get_market_memory`
`memory:write`	Change that knowledge directly. Future searches use the new facts.	No tools yet
`missions:read`	See your missions and what each one has learned.	`list_missions` `get_mission`
`missions:create`	Start new missions for you.	`create_mission`
`missions:write`	Pause, resume, or edit your existing missions.	No tools yet
`people:save`	Save people to your workspace.	`save_person`
`feedback:write`	Mark results as useful or not and record outcomes.	`record_opportunity_feedback`
`drafts:create`	Write reply drafts. Nothing is sent; you decide what to post.	`draft_response`

Billing, workspace administration and key management are never available to an agent.

You stay in control

Every key appears in your connection list with its scopes, and you can revoke it at any time. Agent activity is attributed to the connection that did it.

Agents can read evidence and use the tools you allowed. They cannot post or message anyone. You review and send every reply.

Open agent settings

You'll be asked to sign in first.