theyasked.me

Docs

Claude & MCP

Connect Claude (or any MCP client) to your feedback: ask, log requests from calls, vote for customers, ship.

TheyAsked ships a remote Model Context Protocol server. Connect Claude (claude.ai, Claude Desktop, Claude Code) or any MCP client to one workspace and ask questions like "what do Enterprise customers want most, by revenue?", log a request from a sales call, add a customer's vote, reply, and publish release notes, without leaving the conversation.

  • Server URL: https://theyasked.me/mcp (Streamable HTTP)
  • Read-only URL: https://theyasked.me/mcp/readonly (never exposes a tool that changes data)
  • Plans: Pro and above.

Connect

claude.ai and Claude Desktop. Settings → Connectors → Add custom connector, name it TheyAsked, URL https://theyasked.me/mcp. Sign in to TheyAsked, pick the workspace, choose what Claude may do (see Permissions), then Allow access. Tokens last an hour and refresh for 60 days; removing someone from the workspace revokes their connector at once.

Claude Code. Create an API key in Settings → API & MCP, then:

claude mcp add --transport http theyasked https://theyasked.me/mcp \
  --header "Authorization: Bearer ta_live_…"

Other clients. Any client that speaks Streamable HTTP with OAuth 2.1 (dynamic client registration, PKCE) or a static Authorization: Bearer header works. npx mcp-remote https://theyasked.me/mcp bridges stdio-only clients.

Permissions

Every connection carries scopes. The OAuth consent screen has one checkbox per scope; API keys get the same picker in Settings → API & MCP. Claude only sees the tools its scopes allow.

Scope Lets Claude
posts:read (always on) Search and read requests, boards, statuses, tags and the roadmap
people:read See customers and companies with plan and MRR, voters, who asked
changelog:read Read changelog entries and drafts
insights:read Rankings by demand and revenue, KPIs, the activity feed
posts:write Create requests for customers, edit them, change status, merge duplicates
votes:write Add or remove votes on behalf of customers
comments:write Post replies and internal notes
tags:write Create tags, tag and untag requests
changelog:write Draft, edit and publish changelog entries
people:write Create or update customers and companies

Older connections with read / write keep working: read covers every read scope, write everything.

Tools

Read (18): search_posts, get_post, top_requests, insights_summary, who_asked_for, list_companies, get_company, list_boards_statuses_tags, list_changelog, find_similar_posts, get_roadmap, list_voters, list_tags, list_vote_lists, list_users, get_user, activity_feed, render_chart.

Write (18): create_post, update_post, update_post_status, merge_posts, vote_on_behalf, remove_vote, add_comment, add_internal_note, create_tag, tag_post, untag_post, create_changelog_draft, update_changelog, publish_changelog, create_vote_list, update_vote_list, upsert_user, upsert_company.

In claude.ai and Claude Desktop several tools also draw an interactive view: ranked requests with sparklines, the feedback briefing, who asked, request and account cards, charts, similar requests and the request you just logged (with confirmed "Add vote" and "Merge" buttons).

Prompts

Slash commands in Claude Code and Claude Desktop: weekly_digest, triage_new_requests, prep_roadmap_review, draft_release_notes, account_brief. Each one tells Claude which tools to call and what to produce; anything that changes data still asks you first.

Safety

  • Your words, not your customers'. Titles, bodies, comments and notes are written by customers. Results mark them as untrusted data and the server tells Claude never to follow instructions found inside them.
  • Confirmation. Every tool is labeled read-only or not; Claude asks before writes. Irreversible tools (merge_posts, remove_vote) also require an explicit confirm: true.
  • Nobody is emailed by accident. Replies and changelog publishing notify customers only when you ask (notify defaults to false); status changes say how many followers were notified.
  • Audit trail. Every change made through MCP is logged with who, through which key or connector, the tool and its target: Settings → API & MCP → AI activity.
  • Limits. 120 requests and 60 writes per minute per connection. Publishing the changelog needs the admin role.

Privacy

The MCP server reads and writes the data of the one workspace you connect, as you, and nothing else. TheyAsked does not see your conversations with Claude: only the tool calls Claude makes and their arguments reach us, and we keep the audit trail of changes (not reads) for 400 days. See the privacy policy and terms. Questions: hello@theyasked.me.