Give your AI agent the context that only people have. Your agent writes the brief, AI Scouts holds a short voice conversation with each person, and your agent reads back what they said.
AI Scouts runs interviews for you. You send one link. An AI interviewer, called a scout, has a spoken conversation with each person who opens it and follows up on their answers the way a good consultant would. You get back a transcript and a summary for each person, and a debrief across everyone.
It is for the knowledge that sits in people’s heads and never gets written down: what a team already knows about a client before a project starts, how experts make a judgement call, why customers do what they do.
An agent that needs to know something from people, and cannot find it in a document, can ask them. A typical run:
create_scout with what it wants to learn, who the participants are and a deadline.invite_participants once you confirm the addresses.get_responses and get_transcript as conversations finish, then run_debrief for the synthesis.Examples: gathering what colleagues know before a client project, collecting how senior staff make a decision so it can be written into guidance or a prompt, interviewing stakeholders before a workshop, and asking users why they dropped out of a flow.
create_scout: Create a scout from a brief: what you want to learn, from whom, and by when. Returns the link to share.invite_participants: Email named people their own link. This sends real email, so the agent is told to confirm the addresses with you first.list_scouts: List your scouts with their status, share link and response counts.get_scout: Read one scout: its brief, deadline, opening question, participants and latest debrief.get_responses: Read each participant's summary, objective by objective.get_transcript: Read the full conversation with one participant, including one still in progress.run_debrief: Synthesise the completed conversations into a verdict per objective, with confidence, dissent and gaps.update_scout: Change a scout's brief, deadline or opening question.reopen_scout: Put a closed or archived scout back in front of people.delete_scout: Permanently delete a scout and everything attached to it.A brief is checked before a scout goes live. If it does not say who the participants are or what decision the answers inform, the server replies with the questions to ask you, and nothing is created.
Server URL
https://myaiscouts.com/api/mcp
Streamable HTTP. One URL for every client.
Claude (web, desktop and mobile)
Open Settings, then Connectors, then Add custom connector. Paste the URL, save, and click Connect. You sign in to AI Scouts and approve access. No key is needed.
Claude Code
claude mcp add --transport http scouts https://myaiscouts.com/api/mcp \ --header "Authorization: Bearer sk_scouts_..."
Cursor and other MCP clients
{
"mcpServers": {
"scouts": {
"url": "https://myaiscouts.com/api/mcp",
"headers": { "Authorization": "Bearer sk_scouts_..." }
}
}
}Keys start with sk_scouts_ and are created in Settings once you have an account. The same key works as x-api-key against the REST API at /api/v1/scouts.
The MCP server is included in every plan, including the free one. The free plan covers one live scout and 25 responses a month. Pro is $69 a month for five live scouts and 250 responses a month. Inviting people is not capped on either plan.
Create a free account to get a key or connect from Claude.
Participants are told they are talking to an AI and agree before the conversation starts. Transcripts belong to the account that owns the scout, and an agent connected to that account can read only that account’s scouts. See the privacy policy.