Discord MCP: community context with moderation kept human

A Discord MCP server lets an MCP-capable agent retrieve permitted Discord messages through a bot and use them as context. Community projects such as v-3/discordmcp and SaseQ/discord-mcp also expose actions, depending on the implementation. For community work, start with summaries and reply drafts; keep posts, deletions and bans in human hands.

What it can do

Start with a question a community manager can verify against the original conversation. Useful pilot tasks include:

• A discussion digest: group recurring onboarding questions, attach message references and identify unresolved points.
• A reply draft: prepare an answer using an approved help document and the relevant discussion; flag unsupported claims for review.
• A moderator handover: describe a disputed exchange, separate reported allegations from observed messages, and leave the decision to a moderator.

These are proposed workflows, not built-in summarization features of every server. The connector retrieves context; your agent and instructions determine the draft. Ask for channel identifiers, message references and the time span covered, so a reviewer can inspect the evidence.

Choose a small task from AI agent first tasks for a small business. Other app connectors belong in MCP servers for apps.

Which Discord MCP servers exist?

Examples available as of 30 Oct 2026 include the following community implementations. These are project repositories, not Discord-maintained documentation or endorsements.

Tool names differ between projects: v-3 documents read-messages, while SaseQ documents read_messages. Copy configuration from the chosen repository, not from an unrelated tutorial. Repository availability does not establish maintenance quality, suitability or successful operation in your environment.

How to connect it safely

1. Define the permitted context and output. Choose a test text channel, a discussion interval and a local draft file. Exclude private support cases and sensitive member details from the pilot. Agree who may review the source messages and what may leave Discord. Use the MCP security checklist to record these decisions.

2. Create a dedicated application bot. Use the Discord Developer Portal and authorize it into the test server. Discord documents the bot OAuth2 scope for bot authorization and authenticates bot users with an application bot token. Do not automate a normal user account; see Discord's bot authorization documentation. Keep the token out of prompts, committed configuration and screenshots.

3. Limit effective permissions. For text-channel history, Discord requires VIEW_CHANNEL and READ_MESSAGE_HISTORY; see the message endpoint documentation, as of 30 Oct 2026. Withhold SEND_MESSAGES, MANAGE_MESSAGES, BAN_MEMBERS and other mutation permissions for this pilot. Avoid ADMINISTRATOR, which bypasses channel overwrites. Inspect inherited roles and channel overwrites using Discord's permissions reference. Test the resulting access rather than trusting the invite settings alone.

4. Check intents separately. Review MESSAGE_CONTENT access for ordinary message bodies. Enable necessary privileged intents in the Developer Portal and match the connector's Gateway configuration; approval may be required. Content restrictions also affect HTTP responses. Follow Discord's Gateway documentation, as of 30 Oct 2026. Permissions and intents solve different problems.

5. Review and configure the connector. Follow its current runtime instructions and inspect the launch command before execution. Require a deployment that exposes only the retrieval tools you need; if necessary, put a reviewed server-side restriction in front of it. Do not assume a project offers a read-only switch. MCP's security guidance recommends restricted local execution and access controls for local HTTP servers. A default server identifier is not an authorization boundary.

6. Connect the editor client. In a supported VS Code agent setup, run MCP: Add Server from the Command Palette. Choose Command (stdio) or HTTP to match the chosen implementation, then review trust and authentication prompts. Client features depend on the selected harness and organizational settings; follow the current VS Code instructions, as of 30 Oct 2026, and the MCP in VS Code guide. Keep write tools unavailable in the client as an additional control.

7. Verify a read before requesting a summary. Retrieve a known test message and inspect the actual tool result. Check that a private test channel remains inaccessible. Then request a digest with evidence references and unanswered questions. A person checks the draft and performs any posting or moderation directly in Discord. If the connector cannot enforce the required boundaries, use an authorized, redacted transcript instead.

When this does not work

Message bodies are blank. Check MESSAGE_CONTENT access before expanding permissions. Some message contexts have exceptions; success there does not prove general access. Consult the Gateway reference.

History is missing. Check channel access and history permissions against the message API. Also ask which interval the connector actually retrieved. A recent-message sample cannot support a claim about the entire community's opinion. Mark incomplete coverage in the draft.

The channel contains instructions aimed at the agent. Treat messages and attachments as evidence, not operational authority. For this workflow, a member's request to export another channel or delete a post must never expand the task. Keep secrets out of the context and require review of links before opening them.

You need automatic enforcement. This workflow intentionally stops at a recommendation. If your requirement is unattended bans, deletion or posting, define and review a separate system before granting those capabilities. More permissions are not a remedy for an inaccurate summary.

Where Tvijo fits

As of 30 Oct 2026, use Tvijo here for work on approved community context: preparing a digest, drafting replies or reviewing a proposed response. The DSHREM VS Code extension provides an editor entry point for questions and attached context. A practical starting point is a redacted transcript you deliberately supply, followed by a human review of the draft. Read permissions and data flow before sending member content.

TvijoCode is a test build based on Code - OSS with the extension embedded. It is an editor option, not evidence of a ready Discord integration. Tvijo MCP makes Tvijo callable from a compatible MCP client for consultation; supplied-diff review is prepared functionality, with publication still pending in the dated product update. Editor MCP integration and safe delegation remain planned work. Connecting Tvijo MCP does not automatically give the Tvijo extension access to a Discord MCP server.

At the Tvijo workspace, as of 30 Oct 2026, Try costs $0: sponsored usage ends at 5 tasks, 50,000 tokens or $0.05, whichever is reached first, and draws from a shared $1 daily pool funded by TVIJO. Service credit is shown as $10 for requesting prepaid credit: usage is metered with an 8% service markup. Payment collection is not activated; creating an order is not a charge.

Keep the deliverable reviewable: source references, a draft and explicit uncertainty. A human remains responsible for posts, deletions and bans.

FAQ

Is a Discord AI bot the same as a Discord MCP server?

They describe different roles. In the repositories above, the MCP server exposes Discord operations to an external agent through a bot. Choose the integration around the task you need, not the label alone.

What should the first pilot deliver?

An evidence-linked digest from an agreed channel and interval, plus a reply draft. Ask a community manager to verify both before widening access.

Can an agent's summary justify a ban?

Treat it as a review aid. A moderator should inspect the underlying messages, apply community rules and make the decision personally.

Discord is a trademark of its owner. Tvijo is not affiliated with or endorsed by it.