Work with Ducalis from
Claude, Codex or ChatGPT.
Bring customer feedback and backlog context into your AI assistant. Connect through cloud MCP with OAuth, turn a request into a reviewed draft and verify the result in Ducalis.
Connect your assistantConnect with OAuth. Start with a read.
Ducalis’s cloud MCP server gives a compatible assistant access to product context and tools available to your account. Choose a client below, then connect using your own account’s server URL.
- In Ducalis, open personal settings → API, MCP.
- Copy Server URL under Connect MCP.
- Add that URL in your client’s remote HTTP MCP connection flow.
- Sign in to Ducalis in the browser, review the requested access and approve the OAuth connection.
- Return to your client. Ask for a read of one selected board and check the returned sources before changing anything.
Manage approved clients in your API, MCP settings and disconnect a client there when needed. The personal REST API token is separate; OAuth does not require pasting it or your model-provider API key.
┌─ CHECK YOUR FIRST READ ──────────┐ │ Read the selected board │ │ Summarize linked feedback │ ├──────────────────────────────────┤ │ Do not change any records │ │ Check returned source links │ │ Then choose a product task │ └──────────────────────────────────┘
Connect Claude, Codex or ChatGPT
Use the Server URL shown in your Ducalis account. Client menus and availability depend on your version, plan and workspace settings.
Claude
In Claude, open Customize → Connectors and add a custom remote connector named Ducalis. Paste your Server URL, follow the sign-in flow and enable the connector for the conversation. Team workspaces may need an owner to add it first.
For Claude Code, add the URL as a remote HTTP MCP server. Use /mcp to authenticate in the browser and check the connection.
Codex
In the Codex IDE extension, open the gear menu → MCP servers → Add server. Choose Streamable HTTP and paste your Ducalis Server URL. Save, restart the extension and choose Authenticate to complete OAuth.
In Codex CLI, add Ducalis with codex mcp add ducalis --url <YOUR_SERVER_URL>, then run codex mcp login ducalis. Use /mcp to inspect active servers.
ChatGPT
Open ChatGPT Plugins, choose the plus button and Add custom MCP server. Enter your Ducalis Server URL and choose OAuth authentication. Review the setup and create the custom plugin, then install it and sign in to Ducalis when prompted.
Enable its tools in your conversation and try a scoped read. Your workspace’s plugin controls can affect availability.
ChatGPT MCP connection guide →Turn feedback into requirements you can review
Start with one feature request rather than your entire workspace. Ask your assistant to read its source feedback and linked backlog work, then draft the problem, user story and acceptance criteria.
Try a read first: “Read the invoice-export request on my selected board. Summarize the feedback and linked issues with source links. Do not change any records.”
Then request a draft: “Draft requirements, a user story and acceptance criteria from this feedback. Mark missing information as open questions. Show the proposal before saving anything.”
Check the evidence and scope with your team. Once you approve the text and destination, ask the assistant to apply the supported change using the available Ducalis tools. Open the returned link in Ducalis and verify the saved result.
┌─ VERIFY IN DUCALIS ──────────────┐ │ Open the returned issue link │ │ Check saved text and sources │ │ Confirm the intended board │ ├──────────────────────────────────┤ │ Reviewed work in your backlog │ │ Team chooses the next priority │ └──────────────────────────────────┘
One product workflow, your choice of assistant
Use the context in Ducalis to prepare a decision, then review what changes.
Understand feedback
Ask for related requests, source evidence and gaps in the information. Check proposed duplicates before merging them.
Organize feature requests →Review backlog priorities
Discuss impact, effort and customer context. Review suggested scores with the team before deciding what belongs in the roadmap.
Prioritize the backlog →Prepare a release update
Draft an update from selected completed work. Check the customer-facing wording and approve it before publication.
Automate release notes →Ducalis access and the tools your client exposes determine the available actions. Client confirmation settings vary: explicitly request a draft when you want to review before a write. A successful sign-in is the start; verify the result of each intended action.
MCP connects your assistant. BYOK powers built-in AI.
These are separate settings and can coexist.
Cloud MCP + OAuth
Your external client provides its model access. OAuth authorizes that client to work with Ducalis context and tools. Manage the connection under personal API, MCP settings. Ducalis runs in the cloud; use your phone when your chosen client supports remote MCP there.
Connect your assistant →Bring your own model key
For Ducalis’s built-in AI Assistant and messenger bot, configure Provider, optional Base URL, API Key and Default Model under Integration, Billing → AI → AI provider. Connecting an external MCP client keeps this organization setting in place.
Explore BYOK and built-in chat →You can also start directly in Ducalis’s built-in AI chat. Choose the entry point that fits your workflow.
Choose a workflow for your connected assistant
OAuth connection steps are above. These resources help you choose a task and review the product work you do with your assistant.
Start with one customer request.
Connect your assistant, check the context, review the draft and bring approved work into your feedback-to-release workflow.
Start free