Your agent can create and update Productboard content as well as read it. From a conversation in Claude, Cursor, or another connected coding agent, you can create features and initiatives, fill in fields, score work with your custom driver fields, and link items together to keep your product hierarchy meaningful.
An agent could read your product hierarchy, but creating or changing anything meant stepping back into Productboard and doing it by hand. So workflows that used to stop at the point of creation can now run end to end.
Note: This guide assumes your agent is already connected to Productboard. If it isn't, start with Connect a coding agent to Spark via MCP.
In this article:
- What you can create and edit
- Creating an initiative and linking it up
- Building out a hierarchy: linking features to an initiative
- Scoring with driver fields
- Working with companies and users
- Posting health updates
- Managing portal cards
- Organizing documents and folders
- Good habits when your agent writes to Productboard
- See also
What you can create and edit
Your agent connects as you, so it can only create and change what your Productboard role allows. Within those permissions, it can:
- Create and update entities: objectives, initiatives, key results, components, features, subfeatures, companies, and users.
- Write to any entity field: names, descriptions, owners, teams, statuses, and custom fields, including driver fields such as Value, Risk, Impact, Confidence, and Effort. @ mentions to Productboard entities in any text field are resolved and persisted correctly, so references link up the same way they do when you type them directly.
- Manage relationships: link features to initiatives, connect initiatives to objectives, and update how items relate as your plans change.
- Read relationships: ask for a complete picture of how items connect. For example, which features roll up to which initiative, and which initiatives support which objective.
- Organize documents and folders: create documents and folders, and list, move, and rename them. See Organizing documents and folders.
- Post health updates: post a health status and update text on objectives, key results, initiatives, and features. See Posting health updates.
- Manage portal cards: create, update, publish, move, and unpublish a feature's portal card. See Managing portal cards.
Creating an initiative and linking it up
Start with plain language. Name the item, say who owns it, and mention any fields you want filled in, for example:
- "Use Productboard MCP to create an initiative called Onboarding Revamp. Make me the owner and assign the team to Growth."
Then connect it to the objectives it supports. You can refer to objectives by name, by ID, or by pasting a Productboard link; all three work:
- "Link the initiative to the following objectives: Reduce time to first value and Increase activation rate."
Your agent confirms what it created, and the new initiative appears in your workspace immediately, with the owner, team, and objective links in place. Refresh your board and it's there.
Building out a hierarchy: linking features to an initiative
An initiative is more meaningful with its features attached. Gather the features that belong, by name or by copying their links straight from Productboard, and hand them over in one message:
- "Link the following features to the initiative: Accept/reject changes, Version history for edits, and [pasted Productboard link]."
Two things make this feel effortless:
- Context carries over. If you just created the initiative in the same conversation, you don't need to name it again. "The initiative" is enough.
- Links resolve automatically. Paste a Productboard URL and the agent works out what it points to.
Because your agent can also read relationships, you can check the result without leaving the conversation: "Show me everything linked to the Onboarding Revamp initiative."
Scoring with driver fields
Driver fields such as Value, Risk, Impact, Confidence, and Effort are now writable by your agent. That makes scoring automations like RICE possible, because the scores no longer have to be entered by hand at the end. Try a prompt like:
- "For each feature under the Onboarding Revamp initiative, review the linked customer feedback and feature spec, then use the available evidence to fill in the Value, Risk, Impact, Confidence, and Effort fields. After scoring each feature, add a short justification for each score as a comment in the Scoring section of the feature spec."
The agent does the analysis and writes the scores into the same fields your prioritization boards already use, so there's no copying numbers from a chat window into Productboard by hand.
Working with companies and users
Agents can also create and update company and user records, so customer context can flow into Productboard as part of an agent workflow. For example, when an agent processes feedback that mentions a customer who isn't in your workspace yet, it can create the company record itself instead of stopping to ask you to add it.
Posting health updates
Your agent can read an item's health updates and post new ones on objectives, key results, initiatives, and features. Each update has a health status (On track, At risk, or Off track) and update text. A health update posted by your agent works the same as one you post in Productboard: followers get an email notification, and the status rolls up to parent items.
You need edit access to an item to post a health update on it. Try prompts like:
- "Use Productboard MCP to show me the health updates for the Onboarding Revamp initiative."
- "Use Productboard MCP to mark the Onboarding Revamp initiative as At risk. Update text: blocked on API review, new ETA is November 30."
Note: Your agent can post new health updates but can't edit or delete existing ones. Edit or delete them in Productboard. If an item's health was rolled up from its child items, a posted update switches it to a manual health status. See Overwriting the automatic roll-up.
Managing portal cards
Your agent can work with the portal card for any feature you can edit. Within your permissions, it can:
- Create a card: create a portal card for a feature that doesn't have one yet.
- Read a card: see a card's name and description, and the portals, tabs, and sections it appears on.
- Update a card: change a card's name or description.
- Publish a card: place a card on a portal, including a card you unpublished earlier.
- Move a card: change the tab or section a card appears in on a portal.
- Unpublish a card: remove a card from a portal. The card itself isn't deleted.
Try prompts like:
- "Use Productboard MCP to read the portal card for the Bulk owner editing feature."
- "Rewrite the portal card description for Bulk owner editing based on its spec. Show me the text before you save it."
- "Move the Bulk owner editing card to the Launched tab on the Customer Feedback portal."
- "Unpublish the Bulk owner editing card from the Customer Feedback portal."
Note: Customers see card text as soon as a card is on a portal. Review the wording your agent proposes before it saves. Your agent can't change card images or post portal card updates.
Organizing documents and folders
Your agent can work across your whole folder structure, not only your personal space. It can look inside Agent Knowledge, teamspaces, and folders, then create, file, move, and rename documents and folders there. Workflows that end with filing a document, such as adding research to Agent Knowledge, can run end to end.
Within your permissions, your agent can:
- List folder contents: see what's directly inside Agent Knowledge, a teamspace, a folder, or a personal space. Results go one level at a time and are split into pages. Each item shows its title, type, URL, and last-updated time.
- Create documents: create a document in any supported location. If you don't name a destination, the document goes to your personal space.
- Create folders: create an empty folder in any supported location.
- Move documents and folders: move a document, or a folder along with everything inside it. IDs, URLs, and comments stay the same after the move.
- Rename documents and folders: give a document or folder a new name of up to 80 characters.
Point your agent at a specific document or folder by pasting its Productboard link or giving its ID. An ID that comes back from a folder listing works as-is in any other request. Try prompts like:
- "Use Productboard MCP to list what's in the Research folder in the Growth teamspace."
- "Find the latest weekly update in the Weekly updates folder and summarize the open risks."
- "Save the competitor summary we just wrote as a document called Competitor pricing in Agent Knowledge."
- "Create a folder called Archive in the Growth teamspace and move every document with 'draft' in its title into it."
Note: Boards and entity specs can't be moved.
Good habits when your agent writes to Productboard
- Start requests with "use Productboard MCP…" so your agent reaches for the connection.
- Check for existing work first. Before creating something new, ask your agent to search for related items. Your search will match on meaning as well as keywords, so it finds related work even when the names differ. You can also narrow it by fields such as status or owner. For example: "Use Productboard MCP to check whether we already have a feature for bulk-editing owners before creating one."
- Review what the agent reports back. After a write, the agent summarizes what it created or changed and usually gives you a link. Click through the first few times to confirm the changes look right.
- Remember it's real data. New items, field changes, and links are immediately visible to teammates and may trigger notifications. Treat agent writes like your own edits.
- Check version history if something looks off. Every MCP write is logged in version history, so you can see exactly what changed and when. Open the item in Productboard and look for the version history control on the spec or document. To see field changes such as status or owner, ask your agent: "Use Productboard MCP to show what changed on the Onboarding Revamp initiative this week."
- You stay in control of access. A workspace admin can turn MCP access off at any time, which immediately ends any connected agent's session. And an agent can never touch content your own role can't.