Listen Labs
by Listen Labs
Browse AI-moderated interview studies, read respondent transcripts and pull analysis from an AI agent. Anthropic lists 5 read tools; Listen Labs documents 11, including one that launches recruitment and spends credits.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Listen Labs via MCP
https://listenlabs.ai/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Listen Labs Tools & Capabilities (11)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- No OAuth scopes exist at all. The server's own RFC 9728 descriptor at
https://mcp.listenlabs.ai/.well-known/oauth-protected-resource/mcpand its authorization server metadata both declaredscopes_supportedas[]on 2026-08-22. The boundary that matters here — between reading a transcript and spending credits to recruit strangers — does not exist at the grant. It exists only in your Listen Labs role, which the consent screen does not let you narrow for this session. - Anthropic's directory understates the surface. It lists 5 tools and records permissions as
Read. Listen Labs names 11 and documents write, launch and delete capabilities. Do not size the risk from the listing. - The real tool count is above 11. Folder create, rename, move and delete are documented capabilities with no published tool name, so we cannot enumerate them and neither can you.
- We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail or
readOnlyHintis published here. Every read/write label on this page is our classification of names and documented behaviour. - Launching costs money and cannot be undone by unlaunching. Listen Labs states a launched study begins collecting responses immediately, that started responses are allowed to finish even past a response limit, and that participant incentives are the customer's responsibility, billed separately from platform credits. You can pause or close a study, but interviews already completed are already paid for.
- Analysis needs a completed run.
get_study_analysisreturns nothing useful until one exists; checkhas_analysisonlist_studiesfirst. - Edits to a live study are invisible until published. Listen Labs states edits stay in draft, and that deleting questions that already have responses removes those answers from analysis.
- Credits cannot be topped up over MCP. Listen Labs' research skill states a short balance has to be resolved in the Listen dashboard.
- Transcript content is untrusted input. It is written by recruited members of the public. Pairing broad transcript reads with
edit_studyorlaunch_studyin one session is the combination to avoid. - No vendor-published rate limits. We found none documented for the MCP server.
Frequently asked questions
Yes. Listen Labs documents a launch_study tool that publishes a study and starts recruitment, and its own research skill marks that call as spending credits. Recruitment draws on a panel Listen Labs describes as 30 million-plus verified respondents in 45-plus countries, so a launch dispatches AI-moderated interviews to real human participants.
The directory entry predates the write tools, or omits them. Anthropic lists five read tools and marks the connector permissions as Read. Listen Labs' own research skill names eleven tools at commit cbff3cf, and its MCP documentation describes creating, editing, launching and deleting. Treat five as a stale floor, not the surface.
None that narrow anything. The server's RFC 9728 resource descriptor and its authorization server metadata both declared scopes_supported as an empty array on 2026-08-22. There is no scope separating reading a transcript from launching recruitment, so consent is all-or-nothing within whatever your Listen Labs user account can already reach.
Interview transcripts of real research participants, and the media behind them. Listen Labs documents transcripts as the complete text of each interview, and its response API returns signed audio URLs valid for one hour plus HLS and MP4 video URLs. Screen-share studies also capture recorded application windows or full screens.
Yes, folders. Listen Labs' MCP permissions state the integration can create, rename, move or delete folders, and its troubleshooting page confirms only empty folders can be deleted. No documented tool deletes a study or a response. Listen Labs publishes no tool name for any folder operation, so the delete path is unnamed.
Your account role and a client-side gate, not the connector. Listen Labs defines a Collaborator role that can build studies but cannot launch them, and only Researchers and Admins can start recruitment. Listen Labs' own research skill separately instructs the agent to never launch on its own initiative and to require explicit confirmation.
The study has no completed analysis run yet. Listen Labs states that get_study_analysis requires a finished analysis, and that new or in-progress studies may not have one. Check list_studies for has_analysis set to true before asking for a report. Reports update continuously as new responses arrive.
The client is using the wrong transport. Listen Labs states the server is stateless and accepts only POST on the /mcp path, and that clients must use Streamable HTTP rather than SSE session polling on GET. Confirmed live on 2026-08-22: a GET to the endpoint returns 401 rather than a session stream.
Sources
- Listen Labs MCP documentation index (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs, "Connect Your MCP Client" (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs, "Troubleshooting & Reference" (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs documentation index,
llms.txtandllms-full.txt— · (retrieved 2026-08-22).docs.listenlabs.airedirects todocs.listenlabs.com; the directory's documentation URL resolves here. · retrieved 2026-08-22 - Listen Labs research skill,
SKILL.mdat commitcbff3cf(retrieved 2026-08-22). Linked from Listen Labs' own connect page as their sanctioned skill; the source of the 11 tool names. · retrieved 2026-08-22 - Listen Labs, "Recruiting 101" (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs, "Responses Tab & Transcripts" (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs, "Seat Types & Permissions" (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs, "Frequently Asked Questions" (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs API reference, Launch Study and List Wallets — · (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs API reference, Get Single Response (retrieved 2026-08-22) · retrieved 2026-08-22
- Listen Labs, "Email Campaign" (retrieved 2026-08-22) · retrieved 2026-08-22
- Live OAuth posture check: anonymous
initializetohttps://listenlabs.ai/mcpandhttps://mcp.listenlabs.ai/mcpreturned 401 with awww-authenticateBearer challenge; RFC 9728 metadata athttps://mcp.listenlabs.ai/.well-known/oauth-protected-resource/mcpand authorization server metadata athttps://mcp.listenlabs.ai/.well-known/oauth-authorization-server(2026-08-22) · retrieved 2026-08-22 docs.listenlabs.com/robots.txtcarriesContent-Signal: ai-train=yes, search=yes, ai-input=yes— synthesis expressly permitted by the publisher (retrieved 2026-08-22) · retrieved 2026-08-22- Anthropic Connectors Directory entry — , read from our directory snapshot dated 2026-08-16 · retrieved 2026-08-16
- Listen Labs support — <mailto:support@listenlabs.ai> · Privacy (the directory publishes
https://listenlabs.ai/privacy, which redirects here)
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Productivity
- Developer
- Listen Labs
- Tools
- 11
- Domain
- listenlabs.ai
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.