Model Context Protocol
Let an agent grow you. Without the keys to post.
Quillen speaks MCP, so any agent that does — Claude, Cursor, your own — can find your prospects, score buyer intent, read your analytics, and draft replies in your voice. Every write lands in your review queue. Nothing it does can post.
The guarantee
Your agent proposes. You dispose.
Every other tool hands an agent your posting credentials and hopes. Quillen doesn’t. The agent can read everything and draft anything, but the one write it can make is a draft into your queue. Wiring an agent to your real account is finally safe, because the agent literally cannot post.
Four tools
Read three. Queue one.
discover_prospectsreadsSearches X and LinkedIn for people who match a description and scores each one for buyer intent against your ICP. The agent asks; Quillen searches and ranks.
list_candidatesreadsReturns the prospects already surfaced into your workspace, newest first, with the post that flagged each one. Free, no new search.
get_analyticsreadsPost performance across X and LinkedIn over a window: engagement, impressions, followers, per-platform split. So the agent can decide what is working.
draft_replyqueuesWrites a reply in your own SOUL.md voice and puts it in your review queue. This is the only tool that changes anything, and all it changes is a draft.
Connect in a minute
Paste one block, swap in your key.
Create a key in Settings, drop this into your MCP client’s config, and your agent can see your workspace. Scope it to exactly what the agent needs; revoke it whenever you like.
{
"mcpServers": {
"quillen": {
"type": "http",
"url": "https://www.quillen.xyz/api/mcp",
"headers": { "Authorization": "Bearer YOUR_KEY" }
}
}
}Why it’s safe
Safety you can point at, not promise.
No posting scope exists
A key can hold read scopes and draft:write. There is no post scope to grant, in the product or the code. A leaked key still cannot publish.
Human approves everything
Every draft an agent queues waits for you in the inbox. Nothing reaches a timeline until a person clicks approve.
Scoped, revocable keys
Mint a key with only the scopes an agent needs, see when it was last used, and revoke it in one click. It breaks the moment you do.
Built for one tenant
A key resolves to exactly one workspace, server side. An agent can never read or touch another customer’s data, whatever it asks for.
Questions
What agents can and cannot do.
Can an AI agent post to my accounts through Quillen?
No. The Quillen MCP server has no posting scope, in the product or the code. An agent can read your analytics and prospects and write a draft into your review queue, but publishing always needs a person to click approve. A leaked key still cannot post.
What can an agent do over the Quillen MCP server?
Four tools. discover_prospects and list_candidates find and score people by buyer intent, get_analytics reads your X and LinkedIn performance, and draft_reply writes a reply in your voice into your review queue. Three read; one queues a draft.
Which AI clients can connect to Quillen?
Any MCP client. Claude, Cursor, or your own agent. Add the Quillen HTTP endpoint and a scoped API key to the client MCP config and the agent can see your workspace.
How do I connect my agent to Quillen over MCP?
Create a scoped API key in Settings, then add the Quillen MCP endpoint at https://www.quillen.xyz/api/mcp with an Authorization Bearer header to your MCP client config. It takes about a minute.
Is it safe to give an AI agent access to my social accounts?
Yes, because Quillen never gives the agent the ability to post. Keys are scoped and revocable, each key resolves to exactly one workspace on the server, and the only write an agent can make is a draft a human must approve.
Point your agent at Quillen.
Create a scoped key and connect in a minute. Free to start, no card, cancel anytime.