Researched reference
A dated, sourced reference of MCP support across eight feedback tools, checked against each vendor's own pages in July 2026. Where we could not verify anything, we say so instead of guessing.
MCP (Model Context Protocol) is an open standard that lets AI coding tools like Claude Code and Cursor call outside services. An MCP server is the service side of that standard: it exposes a set of named tools that an agent can call to read data or make changes. For a feedback tool, that means your coding agent can work directly from your board instead of you copy-pasting requests into your editor.
That is the whole idea in three sentences. If you want the deeper version, with example tool calls and responses, the FeatQ MCP page walks through a full agent session.
A feedback board is a list of things your customers want, ranked by demand. A coding agent is a thing that builds software. MCP is the wire between them, and it closes a loop that used to be manual at every step:
Without MCP, every arrow in that loop is a human copying text between windows. That is why MCP support has quietly become a real evaluation criterion when picking a feedback tool in 2026, next to pricing and voting features. If your team uses Claude Code, Cursor, or any other MCP client daily, a board the agent cannot read is a board that slowly goes stale.
One caution: an agent that can write to your board should be scoped. Check what a connected key can actually change. A key that can touch billing or delete boards is a very different risk from one that can only read requests and update statuses on a single board.
Every row below was checked against the vendor's own pages: homepages, help centers, changelogs, and support articles. "Official" means the vendor builds and supports the server itself. Where nothing verifiable turned up, the row says "no public MCP server found" rather than asserting the tool has none, because vendors ship fast and a page like this can only state what is verifiable today.
Last checked: July 2026
| Tool | MCP server status | Type | Availability | Source |
|---|---|---|---|---|
| FeatQ | Yes. Every board is an MCP server with six tools. | Official | Included on every plan. | featq.com/mcp |
| Canny | Yes. Official MCP server announced on the Canny changelog, working with ChatGPT and Claude. | Official | Limited to teams on the Ideas beta, Pro plan and above, as of July 2026. | Canny changelog |
| Featurebase | Yes. Official MCP server documented in the Featurebase help center: posts, comments, changelogs, and more. | Official | Available on the Professional plan, as of July 2026. | Featurebase help center |
| Productboard | Yes. An official MCP server ships as part of Productboard Spark; community servers built on the public API also exist. | Official | Tied to Productboard Spark, as of July 2026. | Productboard support |
| Sleekplan | Yes. Official MCP server with its own product page: feedback capture, comment replies, changelog publishing, triage. | Official | Included on paid Sleekplan plans, as of July 2026. | sleekplan.com/mcp |
| Upvoty | Yes. Upvoty advertises an MCP integration on its homepage for ChatGPT, Claude, or any MCP-ready client. | Official | Advertised on upvoty.com, as of July 2026. | upvoty.com |
| UserVoice | No public MCP server found as of July 2026. An MCP server is an open request on UserVoice’s own feedback board. | None found | Not applicable. | UserVoice feedback board |
| Frill | No public MCP server found as of July 2026. | None found | Not applicable. | None found |
Canny announced an official MCP server on its changelog, working with ChatGPT and Claude. The catch, as of July 2026, is the gating: the announcement says it is limited to teams using the Ideas beta on the Pro plan and above. So MCP exists, but whether you can use it depends on your plan and on whether your workspace is in that beta. Several community servers built on the Canny API also exist on GitHub if the official one is out of reach.
Featurebase documents an official MCP server in its help center, covering posts, comments, changelogs, help center articles, and support conversations, and it works with any MCP client that reads a standard server config. The help article lists it as available on the Professional plan as of July 2026. Of the larger tools, this is one of the broader official implementations.
Productboard ships MCP as part of Productboard Spark, its newer workspace: coding agents can read product specs and data from Spark, and Spark itself can call out to other MCP servers like Amplitude. If your team is on classic Productboard without Spark, the practical route as of July 2026 is one of the community MCP servers built on the public Productboard API, such as the ones on GitHub with 40 plus tools. Community servers work, but you are trusting a third party's code with your API token, so read it first.
Sleekplan has a dedicated MCP server page describing feedback capture, comment replies, changelog publishing, and triage from inside an editor, included on paid plans as of July 2026. Credit where due: publishing a real product page for the server, with rate limits and hosting details, is more transparency than most vendors offer.
Upvoty's homepage advertises plugging Upvoty "straight into ChatGPT, Claude, or any MCP-ready client", with the AI reading posts, votes, and roadmap status and creating or updating feedback. As of July 2026 we did not find a separate technical documentation page for it, so check with Upvoty for setup details and plan requirements before counting on it.
No public MCP server found as of July 2026. Notably, the request for one sits on UserVoice's own feedback board, where users point out that other tools already have MCP support. UserVoice changed hands in late 2025 and the new owner has stated plans to invest in the product, so this could change; the honest statement today is simply that we could not find one.
No public MCP server found as of July 2026. Frill positions itself as a simple, focused feedback board, and nothing on its public pages or in MCP server directories indicated an official or community server at the time of checking.
Every FeatQ board is an MCP server, on every plan, no beta gate. The server is scoped to one board per key: an agent can read requests and stats, draft specs, and update statuses, and it cannot touch other boards, billing, or settings. The six tools are below.
A note on fairness: this table states what was verifiable in July 2026, from each vendor's own pages. Any of these companies may ship or expand an MCP server after this was checked, and "no public MCP server found" means exactly that, not "this tool will never have one". If you spot something we missed, tell us and we will update the table.
Each tool runs against the one board its key belongs to and returns JSON the agent can inspect. One line each; the MCP page shows full example calls and responses.
list_requestsList the requests on your board, filtered by status and sorted by votes or by newest.
get_requestRead one request in full, including its comments, before planning or changing anything.
generate_specTurn a request and its discussion into a markdown implementation spec the agent can work from.
update_request_statusMove a request through the workflow; marking it done publishes a changelog entry and emails the voters.
get_board_statsA compact overview of the board: totals, status distribution, and the top-voted requests.
get_ranked_requestsThe board ranked by current demand: votes plus half-weight comments, fading with age, so the top item answers "what should I build next?"
For FeatQ it is one command: copy your board key from the board admin page and add https://featq.com/api/mcp as a streamable HTTP server with the key as a bearer header. Step-by-step setup for Claude Code, Cursor, Windsurf, and other clients lives on the integrations pages, and the full tool reference is in the MCP docs.
If you want the loop to run without you in it, FeatQ also exposes plain HTTP endpoints a scheduler can call, so a weekly job can fetch the top-ranked spec and hand it to your agent. The recipes are in the automation guide.
For the other tools in the table, follow the source link in each row: the vendor's own documentation is always the authoritative setup guide, and plan requirements change more often than pages like this one get rechecked.
Practical answers about agents, voting, embeds, and pricing.
As of July 2026: FeatQ, Canny, Featurebase, Productboard, Sleekplan, and Upvoty each offer an official MCP server or MCP integration, though plan requirements differ. For UserVoice and Frill, no public MCP server was found as of July 2026.
An official MCP server is built and supported by the vendor itself, so it tracks the product and gets fixed when the API changes. A community MCP server is built by a third party on top of the vendor’s public API: it can work well, but you depend on a volunteer maintainer and should review the code before giving it an API key.
No public MCP server was found for UserVoice as of July 2026. There is an open request for one on UserVoice’s own feedback board, so that may change; check their documentation for the current status.
Yes. Every FeatQ board is an MCP server on every plan, with six per-board tools: listing and reading requests, a demand-ranked list, spec generation, status updates, and board stats.
For FeatQ: copy the board key from your board admin page, then run one "claude mcp add" command pointing at https://featq.com/api/mcp with the key as a bearer header. The MCP docs and integrations pages walk through Claude Code, Cursor, Windsurf, and other clients step by step.
Still have questions? Contact us