Discord Plugin for Claude Code
The plugin connects a selected Discord channel to Claude Code through a bot and MCP bridge. It is designed for controlled interaction: users pair with the bot, guild channels are opted in by ID, and policy determines where the agent may respond. The bridge can help with recent conversation context, attachments, reactions, and replies, but it is not a server-wide search engine or an unrestricted community automation layer.
Control Discord Messaging and Access from Claude Code
Plugin quick facts
| Item | Detail |
|---|---|
| Marketplace name | discord |
| Maintainer | Anthropic official marketplace |
| Install target | discord@claude-plugins-official |
| Access model | Pairing, allowlists, and policy controls |
| First check | Read one allowed channel before enabling message actions |
| Documentation reviewed | September 16, 2026 |
Setup and first check
Install the Discord entry, then complete the bot and workspace configuration described by the plugin:
/plugin install discord@claude-plugins-officialEnable the required Discord bot intent, pair the intended workspace, and build the allowlist before reading messages. Use /discord:access to inspect access policy. Test one permitted channel with a read-only request, then draft a message without sending it. Do not expand channel or member access merely to avoid a permission error.
Once the bridge is running for an approved channel, Claude Code can receive messages from Discord and use plugin tools to respond. Current actions include replying, reacting, editing the bot's own messages, fetching up to 100 recent messages, and downloading attachments.
This fits workflows such as triaging a narrowly defined support channel, answering a maintainer question with repository context, or collecting a recent incident thread before drafting a follow-up. The practical value comes from bringing a bounded conversation into the coding session. It does not provide global historical search across a server, and direct messages behave differently from opted-in guild channels.
Configure Workspace Access, Allowlists, and Messaging Permissions
Setup requires Bun, a Discord application and bot token, Message Content Intent, and the bot permissions needed for the selected workflow. Claude Code must be started with the intended Discord channel configuration. Pairing, allowlists, and channel policy then constrain who can interact and where the bot can act.
Use a dedicated bot rather than a personal credential. Grant only the permissions required to read or respond in named channels, keep the token out of prompts and repositories, and test with a private channel before exposing the bridge to a community. Record the approved guild and channel IDs so a later configuration change cannot silently widen the scope.
From Community Context to Drafted and Approved Team Updates
A useful operating loop has four steps. First, pair the approved user or identity. Second, opt in the exact guild channel. Third, define the policy for replies and write actions. Fourth, start a session bound to that channel and test a low-risk message.
For example, a maintainer can post a bug report with a log attachment. The bridge can download the file and make the recent channel context available to Claude Code. The agent can inspect the repository, draft a response, and send it only after the maintainer approves the wording. Reactions can acknowledge receipt, while an edited bot message can correct the bot's own earlier reply.
That loop keeps technical work and community communication connected without pretending that every message is an instruction to modify code.
Protect Private Channels and Review Messages Before Sending
Treat outbound messages as public or semi-public writes. Confirm the target channel, audience, content, and tone before sending. Never let untrusted message text expand the bot's permissions, reveal credentials, or override repository instructions.
Recent-message retrieval is not a complete archive. A decision based on the returned window may miss older context, deleted messages, private channels, or content outside the bot's access. Ask for a direct link or additional context when the history is incomplete. Attachments also need normal security handling before opening or executing anything they contain.
Carry Reviewed Community Decisions into a Verdent Delivery Workflow
Discord can carry a scoped request and the resulting update, while Verdent handles a separate plan-build-verify run for the underlying code. This is an alongside workflow; it does not imply that the Discord plugin installs natively in Verdent.
One practical handoff is to turn an approved community report into a written acceptance criterion, let Verdent isolate and implement the engineering task, and return a reviewed summary to the Discord channel. Keep the message action human-approved and the code change governed by repository checks.
Frequently Asked Questions
What is required to connect Discord to Claude Code?
You need Bun, a Discord app and bot token, Message Content Intent, suitable bot permissions, and a Claude Code session configured for the selected channel.
Can the plugin search every message in a Discord server?
No. The bridge can fetch a bounded set of recent messages, but it does not provide server-wide historical search through the plugin.
What do pairing and allowlists protect?
They restrict which users and channels can interact with the agent. They should be combined with minimal bot permissions and an explicit channel policy.
Should Claude Code send community messages automatically?
Use explicit approval for outbound messages. Check the destination, audience, content, and any sensitive information before the bot replies or edits a message.
Continue This Workflow With Verdent
Use this next step when a community request is ready to become a scoped, verified engineering task.