Studio open until 22 October
Model Context Protocol

One MCP layer between AI and your Discord server.

Connect ChatGPT, Claude, Cursor, and MCP-compatible AI assistants to your Discord server. Qreate provides discovery, structured planning, stored proposals, and human approval controls.

ChatGPT / Claude / Cursor
https://mcp.qreate.gg/mcp
OAuth 2.1 PKCE
Stored Proposal
Admin Approval
Discord Execution

The AI client talks. Qreate controls what can actually happen.

External AI tools excel at drafting and reasoning. Qreate acts as the boundary that protects your server: verifying permissions, generating stored plans, and executing only what you approve.

Why Discord management needs an MCP boundary

Granting AI models unrestricted bot tokens exposes communities to prompt injection and irreversible errors.

Uncontrolled API access

Direct bot tokens allow external scripts to modify permissions, delete channels, or mismanage members without oversight.

Prompt injection risks

Malicious user messages in an AI chat could hijack bot commands if execution is not isolated.

Lack of server state checks

External clients do not understand role hierarchies or conflicting module configurations without server-side validation.

20 tools across context, planning, and safe apply

Qreate exposes dedicated MCP tools for reading server context, discovering modules, creating reviewable plans, previewing messages, and applying approved proposals.

Scoped token validation

Tokens require the qreate.apply scope for writes. Without it, clients can only read context and save inactive drafts.

Five allowed change kinds

Supports create, update, configure, message, and recipe changes. Deletions, member moderation, and authority changes are withheld.

Immutable stored plans

Every proposal is hashed and stored server-side. Approvals bind to the plan ID and version, preventing tampering.

Workflow: Claude Desktop server audit and proposal

Connect Claude Desktop to https://mcp.qreate.gg/mcp using OAuth 2.1.
Ask Claude to inspect active server modules and propose an onboarding improvement.
Claude queries context tools and calls qreate_create_plan to draft verification and role settings.
You review the stored proposal card and confirm approval. Qreate executes the verified plan.

Non-negotiable security guarantees

Qreate checks permissions on every request. Connected AI clients cannot widen their scope, delete channels or roles, ban members, or alter Q authority levels.

MCP integration surfaces

Tools available to compatible clients across planning, discovery, and publishing.

Discovery ToolsServer ContextProposal PlanningMessage PreviewSafe ApplyAudit Records

Frequently asked questions

Where is the Qreate MCP endpoint hosted?

The official endpoint is https://mcp.qreate.gg/mcp. It supports OAuth 2.1 with PKCE and dynamic client registration.

Which operations are blocked from MCP clients?

Qreate withholds three change kinds: remove (deletions), member (moderation actions on users), and authority (changing permissions or authority modes).

Can an MCP client apply changes without approval?

No. Planning over MCP never executes automatically. Applying a proposal requires explicit user approval bound to the stored plan ID.

Does the MCP server duplicate authorization logic?

No. The MCP server acts as an adapter, forwarding the user's token to Qreate's API where standard guild administration permissions are verified.

Tell Q what your server needs.

Studio access is open free until 22 October 2026 with no credit card required. A permanent free plan remains available. Paid plans will be announced separately.