Dis.gg MCP Docs

Changelog

This page gives a short overview of public-facing updates to the hosted Dis.gg MCP service.

2026-08-13 #

  • OpenRouter image generation now accepts explicit output requests such as 1920x1080 and lets OpenRouter map them to each model's supported resolution tier, instead of restricting callers to a fixed legacy size list.

2026-08-11 #

  • OpenRouter image generation now uses the dedicated Images API, enabling current image-only models such as Qwen Image 3.
  • Image clients can request validated inline base64 with X-Image-Content: raw and bypass URL uploader storage when they can attach the result directly.
  • Image-provider requests now allow up to 100 seconds for slower generation jobs to finish.
  • Image generation now carries caller-scoped medium risk metadata instead of being grouped with moderation and service-configuration operations.
  • Image generation now accepts bounded reference images for editing. Images from the current message are applied automatically, while a previous generated image requires an explicit refinement choice from the model.
  • Image tool metadata now explicitly tells models to call the tool for image creation or editing and never claim success from a text-only response.

2026-08-05 #

  • Message-reading tools now tell models when message_content is not declared for the current MCP session. This clarifies that Discord will normally redact message-content fields while metadata remains usable, including when fetch_messages is available through server_members.

2026-04-20 #

  • tools/list now returns a tier field on every tool, typically one of user, moderator, manager, or app_owner. Integrators can use it to decide which tools to surface eagerly vs. on demand — for instance, keeping day-to-day tools always available while loading more privileged ones only when needed. Tier is derived from the same access requirements tools/list already enforces; it never grants access the caller did not already have.

2026-04-16 #

  • Improved JSON-RPC error messages so failures are easier to read and triage from common MCP clients.
  • Documentation helper tools now advertise their risk profile in tools/list, alongside the rest of the public tool catalog.

2026-04-15 #

  • X-Privileged-Intents now accepts an empty value as an explicit "no privileged intents" signal. Surrounding whitespace is also tolerated and trimmed before the value is parsed.

2026-04-14 #

  • Added mcp_docs_search, mcp_docs_read, and mcp_docs_list so an application can browse this documentation directly from its MCP session without opening a browser.
  • Webhook retries for reminders and util_report_user events are now encrypted at rest while they wait in the delivery queue.

2026-04-06 #

  • Added mcp_application_stats so application owners can read their own tool call, image generation, and error counters. It hides automatically when the caller already has an operator-level stats view, to avoid exposing the same numbers twice.
  • Renamed util_staff_report to util_report_user. The behavior is unchanged; the new name better reflects that the report targets a specific user rather than "staff".
  • Event delivery for reminders and reports is more resilient: failed deliveries are retried automatically through a dedicated retry queue, and reminder counters stay accurate even under burst load.

2026-04-05 #

  • Unified the landing page description and OpenGraph summary so link previews match the text shown on the page itself.

2026-04-04 #

  • Reminders now default to the current channel when channel_id is not specified. Set channel_id to "0" to deliver via DM instead. The response includes effective_channel_id to confirm where the reminder will be delivered.
  • Added fire_at parameter to util_schedule_reminder as an alternative to delay_seconds, for scheduling at an absolute UNIX timestamp.
  • Tightened how event_webhook tokens are validated when you set them — non-URL-safe characters are now rejected, and extra whitespace is cleaned up automatically.
  • Added mcp_service_feedback so application owners can send quick feedback about the service without leaving their integration.

2026-03-30 #

  • Image generation now checks that the downloaded file is actually a supported image before trying to reupload it.
  • Upload checks are stricter: the provider must return a complete response, and the local uploader verifies the file it just wrote.

2026-03-29 #

  • Reworked how image generation and upload providers are selected, now driven by X-Image-* and X-Upload-* request headers.
  • Renamed the stored image-provider setting to image_provider and dropped the old image-specific service defaults.
  • Image and upload provider overrides sent at request time can now be saved to your app settings, with secret values encrypted at rest.
  • Alibaba image generation no longer enforces a model allowlist.
  • OpenRouter image requests now correctly ask for image output only.
  • Updated provider docs to reflect these changes.

2026-03-26 #

  • util_custom_data values are now encrypted at rest. This tool is still meant for lightweight structured context, but stored data is now better protected.
  • Added OpenRouter as an image-generation provider.
  • Added list_pinned_messages for reading a channel's pinned messages.

2026-03-24 #

  • Fixed tool input schemas so array fields include their item types, making them work with OpenAI-style schema validators.
  • The local upload provider now names files after the message ID ({message_id}-{count}.{ext}) when X-Current-Message is set.
  • Added the event_webhook app setting for receiving event payloads, replacing the previous audit_webhook option. At the time of writing, webhook requests are sent through Cloudflare Workers — if your endpoint is also on Cloudflare Workers, you may need a custom domain since Cloudflare blocks worker-to-worker requests on workers.dev domains. Check the event_webhook section in the MCP tools docs for the current setup.
  • Added scheduled reminder tools: util_schedule_reminder, util_list_channel_reminders, util_list_user_reminders, and util_cancel_reminders, with per-app, per-user, and per-channel limits.
  • Added util_report_user (formerly util_staff_report) so models can report user behavior to your event webhook for review.
  • Reminder and report tools only appear when your app has an event_webhook set.
  • Webhook values are encrypted at rest.

2026-03-23 #

  • Member timeouts now accept a simple duration instead of requiring you to calculate an absolute timestamp.
  • Added X-Current-User and X-Current-Guild headers, so more tools can figure out the right user, guild, or channel without you repeating IDs every time.
  • Added send_direct_message with controls to restrict or disable DM targets.
  • Added automatic DM prefixing for integrations that want clearer attribution when the bot contacts users directly.
  • Added X-Current-Message for more precise context links in DM attribution.
  • Improved header and tool docs for more predictable guidance.

2026-03-22 #

  • Rewrote the Privacy Policy and Terms of Service to be clearer about operator identity, hosting, data handling, and contact details.
  • Unified the service name and link structure across the website and documentation.
  • Public site assets are now self-hosted instead of relying on a separate static host.

2026-03-21 #

  • mcp_get_access_context now always returns the same payload shape regardless of owner access.
  • delete_own_messages is automatically hidden when your permissions already allow message moderation, reducing tool overlap.
  • get_messages_id_range entries now include a creation timestamp.

This changelog is intentionally concise. For setup and day-to-day usage, start with the other documentation pages.

2026-03-20 #

  • Split plain-text sending from embed sending. Polls stay on their own tool.
  • Added stricter controls for mentions and current-channel replies.
  • Improved tool visibility so clients can guide models more clearly.
  • SSE compatibility improvements: tool failures now reach models with useful error details.

2026-03-19 #

  • Added provider-validated image size support for util_generate_image.
  • tools/list is now more context-aware for image generation.

2026-03-17 #

  • Improved channel management reliability, including safer edit and delete flows.
  • Made webhook responses more predictable.
  • Expanded upload and image-generation provider support.
  • Added encryption for stored provider secrets and maintenance tools for key management.
  • Continued cleanup across tool metadata and documentation.

2026-03-13 #

  • Added image generation and upload provider workflows.
  • Published provider documentation.
  • Improved navigation, layout, and readability across the docs.

2026-03-12 #

  • Added better message range helpers.
  • Improved target scoping and message lookup.

2026-03-11 #

  • Added SSE compatibility endpoints for clients that still use streaming transports.
  • Improved error visibility and resolver stability.
  • Expanded privileged-intent handling and related docs.

2026-03-10 #

  • Improved OpenGraph metadata and documentation quality.
  • Added privileged intent controls and better member resolution.
  • Refined moderation and nickname-management flows.

2026-03-09 #

  • Split public documentation from internal project docs.
  • Added a more complete documentation structure and legal pages.
  • Improved the landing page and docs presentation.

2026-03-08 #

  • Added markdown-based hosted documentation.
  • Simplified docs entrypoints and onboarding.

2026-03-06 #

  • Added richer tool metadata, access hints, and preflight validation.
  • Improved name resolution and debugging support.
  • Strengthened target restrictions and owner-only controls.

2026-03-05 #

  • Initial release.