Customer feedback MCP server
Manage your feedback backlog directly from your AI agent.
Every Reqio project ships its own remote MCP server with 20 tools. Connect your AI coding agent once and give it full project management at your permission level: triage requests, reply to conversations, send announcements, configure the widget, and manage the team.
OAuth 2.1 secured · Role-gated · Works with Claude, GPT, and any MCP client
What it is
A remote MCP server with 20 tools, built into every project
The Model Context Protocol (MCP) is an open standard that lets AI coding agents call external tools over a structured protocol. Reqio implements a remote, Streamable-HTTP MCP server for every project. Your AI agent can manage the full project: triage the backlog, reply to private conversations, send project-wide announcements, update widget branding, manage team members, and rename the project, all using the same interface it uses for everything else.
- Remote MCP server, connects over HTTPS, no local process required
- OAuth 2.1 authenticated, one token per project, audience-bound
- 20 tools covering backlog, conversations, announcements, widget config, members, and project
- Role-gated: the agent gets exactly the permissions of the connecting user, nothing more
- Quota-gated per plan, Free, Pro, and Business limits apply
20 MCP tools
What your AI agent can do
The Reqio MCP server exposes 20 tools across six domains. Each tool maps to a real dashboard operation, gated by the connecting user's role. Billing, project deletion, and token issuance are never MCP tools.
list_featuresList the feedback backlog with optional filters for status, category, and sort order. Returns titles, vote counts, and current status.
get_featureGet the full detail of a single feature request: title, status, internal note, category, vote count, and submitter info.
get_project_statsGet aggregate stats for the project: total requests, requests by status, total votes, and requests by category.
change_statusMove a feature request through the pipeline: Needs action, In progress, or Completed. Triggers subscriber notifications on Completed.
set_developer_noteWrite or update the private internal note on a feature request. Useful for leaving implementation context, blockers, or target release info.
add_commentAdd a comment to a feature request. Comments are visible to users who submitted or voted on the request.
delete_commentRemove a comment from a feature request. Requires comments:delete scope.
delete_featureDelete a feature request entirely. Requires features:delete scope. Irreversible.
list_conversationsList the private conversation inbox, filterable by kind (bug, unexpected, feedback, question) and state. Returns unread counts and last-active times.
get_conversationGet a full conversation thread including all messages, structured diagnostic context, and requester info.
reply_conversationPost a team reply into a private conversation. Notifies the requester in-widget and by email.
convert_unexpected_to_featureConvert a private unexpected-behavior report into a public feature request on the backlog. Automatically subscribes the original reporter.
send_announcementSend a project-wide announcement to all contactable project recipients. High-reach and irreversible, requires explicit confirmation.
get_widget_configRead the project's current widget and branding configuration.
update_widget_configUpdate widget and branding settings. Plan-gated fields (custom logo, badge removal) are enforced by the same rules as the dashboard.
list_membersList all project members and pending invites, including roles.
invite_memberInvite a person to the project by email. Owner-only. Invited role is Developer.
remove_memberRemove a member's access to the project. Owner-only. Requires explicit confirmation.
rename_projectRename the project. Reflected immediately in the dashboard and widget header.
OAuth 2.1 authentication
One token per project. Role-gated. Scope-gated.
The Reqio MCP server uses OAuth 2.1 for authentication. You connect your AI agent using the standard OAuth flow: Reqio acts as the Authorization Server, your agent is the client, and tokens are issued per project with a specific audience URL. Every tool call validates the token scope and re-runs a live permission check against the connecting user's current role.
- Tokens are opaque and stored in the database, never exposed in logs
- Each token is audience-bound to one project URL (/p/{projectId})
- 20 tools grouped under 15 granular scopes, role-gated, covering all six tool domains
- Role check runs live on every call, the same permission matrix as the dashboard
- Revoke any connected agent from the Settings MCP tab
Getting connected
Connect your AI agent in three steps
Create a Reqio project
Sign up and create a project. The MCP server is provisioned automatically. No separate setup required.
Connect your AI agent
Copy the MCP server URL from the project settings. Add it to your AI agent's MCP server list. The agent will prompt you for OAuth consent on first connection.
Approve the scopes
You see exactly which scopes the agent is requesting, across backlog, conversations, announcements, widget, members, and project. Approve the ones you want. The agent is connected and ready to manage the project.
Use cases
What AI agents do with your project
Morning backlog triage
Ask your AI agent to read the backlog, summarize the top-voted requests, and flag any new error reports submitted since yesterday. It reads directly from live data, no copy-paste.
Reply to conversations
Point your agent at the private inbox. It reads open conversations, drafts a reply, and posts it using `reply_conversation`. The requester gets notified in-widget and by email.
Ship and announce
After a deploy, ask your agent to mark the matching request Completed and send an announcement to all project contacts. Notification goes out automatically, no dashboard tab required.
Onboard a new developer
Ask your agent to invite a new teammate by email using `invite_member`. The invite is sent immediately. Owner-only, so only your owner-connected agent can do it.
Questions about the Reqio MCP server
What is an MCP server?
MCP (Model Context Protocol) is an open standard developed by Anthropic that lets AI coding agents call external tools using a structured, typed protocol. An MCP server exposes a set of named tools with defined input and output schemas. An AI agent connects to it once (via OAuth) and can then call those tools as part of any conversation or task, without you writing any custom integration code.
Is the Reqio MCP server a local server or a remote server?
Remote. The Reqio MCP server runs as a Streamable-HTTP endpoint on reqio.app. There's no local process to run or port to expose. Your AI agent connects to it over HTTPS using a Bearer token. This means it works with any MCP client that supports remote servers, including Claude Desktop, Claude.ai, and any agent using the MCP SDK.
Which AI coding agents work with the Reqio MCP server?
Any MCP-compatible client works. This includes Claude (Anthropic), GPT-based agents using the MCP SDK, and custom agents built on the MCP open standard. The Reqio server speaks the Streamable-HTTP transport, which is the current standard for remote MCP servers.
What scopes can I grant my AI agent?
Reqio supports 15 granular OAuth scopes covering all 20 MCP tools across six domains: backlog (read), status (write), notes (write), comments (write and delete), features (delete), conversations (read and write), announcements (write), widget (read and write), members (read, write, delete), and project (write). You choose which scopes to grant at connection time and can revoke a connection at any time from the MCP tab in project settings. Destructive and high-impact scopes are listed separately on the consent screen and are never bundled into a single connect action.
Can the AI agent do anything my team members can't?
No. The MCP server re-runs a live permission check on every tool call, the same role-based matrix that governs dashboard access. If a scope requires a capability the connecting user doesn't hold, the call is rejected. A developer agent cannot invite or remove members because those capabilities are owner-only. Billing, project deletion, and token issuance are not tools at all, so no agent can reach them regardless of role.
How many MCP calls can my agent make per day?
The daily MCP call quota depends on your plan. Free plan users have a limited quota; Pro and Business plans have higher limits. Each tool call counts as one MCP call. If you exceed your daily quota, calls return a quota-exceeded error until the next day's quota resets.
Is there a feedback tool with MCP support other than Reqio?
Reqio ships a built-in, production-ready, OAuth 2.1-secured remote MCP server with full project management as a core product feature. MCP, with its typed tool schemas, role-gated access, and native support in AI coding agents, is a different category of integration from Zapier or webhook approaches.
Let your AI agent work the feedback backlog.
OAuth 2.1 secured · Free to start · Works with any MCP client