Metaview
by Metaview
Query job-interview transcripts, AI notes and recruiting data from an AI agent. Anthropic lists 16 tools; Metaview's own workflow docs name 19 more, including one that posts candidates to your ATS. OAuth sign-in, no application scopes.
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 Metaview via MCP
https://mcp.metaview.ai/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Metaview Tools & Capabilities (16)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- The consent screen has nothing to narrow. The server's RFC 9728 descriptor returned
scopes_supportedas an empty array on 2026-08-21, and its authorization server advertises onlyemail,offline_access,openidandprofile. There is no read-only grant, so authorising the connector authorises everything the account can already do. - We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter detail,
readOnlyHintordestructiveHintis published here. The read/write split above is derived from tool names and Metaview's prose. - The directory listing is incomplete. 19 tool names appear in Metaview's documentation and not in Anthropic's listing, including the ATS write. Assume the surface is larger than the 16.
- Not every session has a transcript. Metaview states the bot can join a call and produce nothing — too short, ended early, or a processing failure — and those sessions still appear in search results. Its fix is to filter explicitly to conversations that have a transcript.
- Unclassified calls default to "Job Interview", which skews any analysis in a workspace that records mixed conversation types.
- Volume is expensive. Metaview puts one hour of transcript at roughly 10,000–15,000 tokens and recommends full transcripts only for 1–5 conversations, summaries for 5–20, and structured AI fields beyond 20.
- Candidate identity is unreliable by name. Where no name is on record, Metaview falls back to displaying an email address or phone number as the participant's display name.
- Some capability is plan-gated. Metaview documents Reports as a premium feature, sourcing enrichment as credit-metered and non-refundable, and MCP access as controllable per workspace and separately for non-admins.
- Automatic scorecard posting is not an MCP tool. Metaview's scorecard autofill and automatic ATS submission are Metaview app settings, not part of the MCP surface. We found no MCP tool in either source that creates or submits a scorecard.
Frequently asked questions
What data does the Metaview MCP server expose to an AI assistant?
Job-interview content. Metaview's MCP documentation lists conversation transcripts, AI-generated Notes, conversation metadata such as role, stage and interviewer, and Sourcing data including searches, surfaced candidates and candidate feedback. Its workflow guide adds scorecards and structured fields. Transcripts are recordings of named candidates answering questions, so third-party speech enters the model's context.
Can the Metaview connector write anything into my ATS?
Not among the 16 tools Anthropic lists, but yes among the tools Metaview documents. Its Sourcing Copilot workflow names post_sourcing_candidates_to_ats, describing it as capped at 50 candidates per call, roughly two minutes asynchronous, and irreversible from Metaview, with the agent's reasoning posted as a profile note. That tool is absent from the directory listing.
Can the connector send email to a candidate?
Indirectly, and this is the most consequential thing on the surface. manage_candidate_sequence adds a candidate to a Sequence, and Metaview documents Sequences as automated multi-step outreach sent from your own connected Gmail or Outlook mailbox. The tool name reads as bookkeeping; the effect is real email to a real person from your address.
Does Metaview train AI models on interview transcripts?
No. Metaview's privacy and security overview states plainly that it does not use customer data for AI model training, and that it uses existing pre-trained models from LLM partners to deliver its services. It describes itself as a data processor that does not own the personal data it processes, with customers retaining ownership.
Do candidates know an interview is being recorded?
That is the customer's responsibility, and Metaview publishes guidance rather than enforcing it. It recommends notifying participants in advance, explaining the purpose, and offering an opt-out, and it offers a managed notice form that records responses. Metaview states it does not provide any option to hide the notetaker from participants in a meeting.
What OAuth permissions does the Metaview MCP server request?
None that narrow anything. Reading the server's own RFC 9728 descriptor on 2026-08-21 returned scopes_supported as an empty array, and its authorization server advertises only email, offline_access, openid and profile — identity claims, not application permissions. The consent screen therefore offers no way to grant reading without also granting the writes.
Who in my workspace can connect Metaview to Claude, and what can they see?
Any member an admin permits, and each sees only their own permitted data. Metaview documents two workspace toggles: one enabling the MCP integration at all, one enabling it for non-admins. Access mirrors existing Metaview permissions, so an admin reaches every conversation in the workspace while a member reaches only their own and those shared with them.
Why did the connector label an internal meeting as a job interview?
Because unclassified calls default to that type. Metaview's MCP guide states that when a conversation's type has not been set, the MCP defaults it to Job Interview, so portfolio reviews, exploratory chats and internal syncs can be miscounted. Its suggested fix is to group results by call type and flag classifications that may be unreliable.
Sources
- Metaview MCP integration overview (retrieved 2026-08-21; source of the connection paths, API-key setup, the data-access list, the permissions model and the compliance note. Anthropic's directory publishes this URL with an
.mdxsuffix, which returns HTTP 308 to the docs index rather than to the page; the canonical URL and its.mdvariant are cited here instead) · retrieved 2026-08-21 - Metaview MCP use cases and workflows (retrieved 2026-08-21; source of the transcripts/notes/fields/scorecards description, the token figures, the transcript-filtering and call-type gotchas, and the email-as-identifier guidance) · retrieved 2026-08-21
- Metaview AI Skills overview (retrieved 2026-08-21; source of the responsible-use rules and the statement that some workflows update records or prepare outreach) · retrieved 2026-08-21
- Metaview Sourcing Copilot workflow (retrieved 2026-08-21; source of
post_sourcing_candidates_to_ats,give_sourcing_feedback,update_application_review_icpand the confirmation rules for irreversible and externally-visible actions) · retrieved 2026-08-21 - Metaview privacy and security overview (retrieved 2026-08-21; source of the no-model-training statement, retention, SOC 2, EU AI Act and FEHA notices) · retrieved 2026-08-21
- Metaview privacy configurations and participant notifications — and (retrieved 2026-08-21; source of the recording-notice guidance, opt-out handling and the refusal to hide the notetaker) · retrieved 2026-08-21
- Metaview workspace settings (retrieved 2026-08-21; source of the two MCP toggles, the audio/video recording rules and the scorecard settings) · retrieved 2026-08-21
- Metaview Sequences overview, settings and FAQs — , and (retrieved 2026-08-21; source of the outreach mechanics, own-mailbox sending, ATS Bcc and sequence-activity posting) · retrieved 2026-08-21
- Metaview
llms.txtandllms-full.txt(retrieved 2026-08-21; the machine-readable index used to locate the MCP pages without crawling the host) · retrieved 2026-08-21 - Metaview
robots.txt(retrieved 2026-08-21; carriesContent-Signal: ai-train=yes, search=yes, ai-input=yes, so synthesis into this page is expressly permitted by the publisher) · retrieved 2026-08-21 - Live OAuth posture check — anonymous
initializetohttps://mcp.metaview.ai/mcpreturned HTTP 401 with an authentication challenge naming the resource descriptor (2026-08-21) · retrieved 2026-08-21 - Live RFC 9728 descriptor — returned HTTP 200 with an empty
scopes_supportedarray (2026-08-21) · retrieved 2026-08-21 - Metaview authorization server metadata (retrieved 2026-08-21; advertises four identity scopes and no application scope) · retrieved 2026-08-21
- Anthropic connector directory (snapshot 2026-08-16; source of the 16
tool_names, the emptyprompt_namesarray, the "Read and write" permissions label, the partner tier, the productivity category and the endpoint) · retrieved 2026-08-16 - Metaview support — · Privacy policy
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
- Metaview
- Tools
- 16
- Domain
- mcp.metaview.ai
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.