Sleekplan

MCP Server

Give Claude, Cursor, Windsurf, VS Code, and any MCP-compatible AI client direct access to your Sleekplan workspace via Streamable HTTP and OAuth 2.1

The Sleekplan MCP server exposes your entire workspace to any compatible AI client. Once connected, you can ask your AI assistant to find feedback, create changelog entries, tag items, look up users, run surveys, and more — all in natural language, without switching to the Sleekplan dashboard. The server is centrally hosted at https://mcp.sleekplan.com/mcp — nothing to install locally.

Before you connect

• The MCP server uses OAuth 2.1 with dynamic client registration. On first connection, your AI client opens a browser window where you sign in to your Sleekplan account and select the workspace to authorize. Subsequent connections are authenticated automatically.
• The server supports Streamable HTTP transport. Clients that only support the older SSE transport can use mcp-remote as a bridge (see the Others tab below).
• Each connection is scoped to a single workspace. To connect multiple workspaces, add a separate MCP server entry for each one.

---

Setup

Open Claude settings, go to **Connectors**, and add Sleekplan. Alternatively, add the following to your `claude_desktop_config.json`:

```json theme={"system"}
{
  "mcpServers": {
    "sleekplan": {
      "url": "https://mcp.sleekplan.com/mcp"
    }
  }
}
```

Restart Claude Desktop after saving the file. On the next launch, Claude will prompt you to authorize your Sleekplan account.

Run the following command in your terminal:

```bash theme={"system"}
claude mcp add sleekplan --transport streamable-http https://mcp.sleekplan.com/mcp
```

Then open a Claude Code session and run `/mcp` to complete the OAuth authorization flow.

1. Open Cursor **Settings**. 2. Navigate to **MCP tools**. 3. Click **Add new server** and fill in the fields:

| Field | Value |
| - | - |
| Name | `Sleekplan` |
| Type | `Streamable HTTP` |
| URL | `https://mcp.sleekplan.com/mcp` |

Save the entry. Cursor will prompt you to authorize the connection on the next use.

Press `Ctrl`/`Cmd` + `P`, type **MCP: Add Server**, select **HTTP**, and enter:

```
https://mcp.sleekplan.com/mcp
```

Alternatively, add the following to your VS Code MCP configuration file:

```json theme={"system"}
{
  "mcpServers": {
    "sleekplan": {
      "url": "https://mcp.sleekplan.com/mcp"
    }
  }
}
```

1. Press `Ctrl`/`Cmd` + `,` to open Windsurf settings. 2. Scroll to **Cascade → MCP servers**. 3. Click **Add Server → Add custom server** and enter:

```json theme={"system"}
{
  "mcpServers": {
    "sleekplan": {
      "url": "https://mcp.sleekplan.com/mcp"
    }
  }
}
```

Save the configuration. Windsurf will prompt you to authenticate on the first use.

Any MCP-compatible client can connect using the server URL directly:

```
https://mcp.sleekplan.com/mcp
```

If your client does not support Streamable HTTP transport, use `mcp-remote` as a compatibility bridge:

```bash theme={"system"}
npx -y mcp-remote https://mcp.sleekplan.com/mcp
```

Note

mcp-remote requires Node.js to be installed. It acts as a local proxy that translates between your client's transport and the Sleekplan server.

---

Available tools

All tools and their parameters are automatically discovered by your AI client — you do not need to configure them manually. Ask your assistant in natural language and it will select the appropriate tool.

Resource	Tools
Feedback	list_feedback, create_feedback, get_feedback, update_feedback, delete_feedback, get_feedback_stats, get_similar_feedback, merge_feedback, tag_feedback
Feedback meta	get_feedback_meta, set_feedback_meta, delete_feedback_meta
Comments	list_comments, create_comment, update_comment, delete_comment
Changelog	list_changelog, create_changelog, get_changelog, update_changelog, delete_changelog
Documents	list_documents, get_document, create_document, update_document, delete_document, link_document_to_feedback, unlink_document_from_feedback, list_document_feedback, list_feedback_documents, list_document_versions, get_document_version, compare_document_versions, restore_document_version
Skills	list_skills, get_skill, create_skill, update_skill
Surveys	list_surveys, get_survey, create_survey, update_survey_name, update_survey_questions, get_survey_summary, get_survey_question_feed, list_survey_responses, get_survey_response
Votes	list_votes, get_voters
Users	list_users, get_user, get_user_segment
Topics	list_topics, list_sub_topics
Tags	list_tags, create_tag, delete_tag
Workspace	list_feedback_types, list_feedback_statuses, list_components, get_category_template, list_admins, list_segments, list_document_types

Documents

Documents hold the thinking behind a request: a plan, a spec, research notes, or the record of why a decision went the way it did. Bodies are Markdown, so an assistant can read and write them directly.

Three things are worth knowing before you prompt against them:

• Documents attach to feedback posts, in both directions. list_feedback_documents answers "what thinking already exists for this request?", which is the useful first call before an assistant works on a post. list_document_feedback answers the reverse, for a document that informs several requests.
• History is append-only. Every save that changes a document adds a version. restore_document_version does not rewind: it writes the older content as a new version on top, so the state being replaced stays recoverable. Use compare_document_versions to see what changed between any two versions.
• update_document is a partial update. Only the fields you pass change, and the rest are left alone. Passing content replaces the whole body, so an assistant should read the document first and send the complete new text.

Ask your assistant for a type key from list_document_types rather than guessing one. Types are workspace-defined, and a workspace can rename them or add its own.

Note

Documents are not available on the Free plan. On a Free workspace, every document tool returns an upgrade message instead of a result.

Feedback meta

Meta is free-form attribution stored on a feedback post: a reporter, a source channel, an account or campaign id. Set it at creation with create_feedback's meta argument, then read and change it afterwards with get_feedback_meta, set_feedback_meta, and delete_feedback_meta. The same keys are filterable through list_feedback.

Two behaviors to expect: values are normalized server-side (spaces become hyphens and other non-alphanumeric characters are stripped), so meta suits slugs and ids rather than free text. And there is no "clear by empty value", so removing a key means calling delete_feedback_meta.

Skills

Skills are reusable instructions your team keeps in Sleekplan under Settings → My Account → Skills: how to triage a bug, how to write a weekly review, which tone a changelog uses. Every connection sees the connected user's own skills plus the ones owners and admins shared with the workspace.

• list_skills returns each skill's slug, name and description. The description says what the skill does and when to use it, so an assistant can match it against the task at hand.
• get_skill takes a slug and returns the skill with its instructions in content (Markdown).
• create_skill saves a new skill from a name, a description and the instructions in content. Assistants call it when you ask them to save how a task was done, or when you accept their offer to save one after a complex task. The new skill is private to you.
• update_skill takes a slug and changes the name, description or instructions of one of your own skills when you ask for it. New instructions replace the old ones, so assistants read the current version first. Skills a teammate shared can't be changed.

Clients that support MCP skills also find every skill as a resource: skill://<slug>/SKILL.md (standard skill frontmatter plus the instructions) and skill://<slug>/_manifest. Tools that read that convention, such as FastMCP's download_skill and sync_skills helpers, can copy your team's skills into a local skills folder.

Deleting a skill and sharing it with your workspace happen on the Skills settings page only.

Note

A skill shared by someone who has since left the workspace, or whose seat was deactivated, is not offered to AI clients. It still appears on the Skills settings page.

---

FAQ

Any MCP-compatible client works with the Sleekplan MCP server. This includes Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, and any other tool that supports the Model Context Protocol. Clients without native Streamable HTTP support can use `mcp-remote` as a bridge. Each MCP connection is scoped to a single workspace, selected during the OAuth authorization flow. To connect more than one workspace, add a separate MCP server entry for each one in your client configuration, giving each a distinct name (for example, `sleekplan-acme` and `sleekplan-beta`). The MCP server uses your Sleekplan account permissions. You can perform the same actions through the MCP server as you can through the Sleekplan admin dashboard — no more, no less. Yes. Authentication uses OAuth 2.1 with PKCE. Your Sleekplan credentials are never shared with the AI client — the client only receives an access token scoped to the workspace you authorized. All communication between your client, the MCP server, and Sleekplan happens over HTTPS. Clear your cached authentication tokens in the AI client and reconnect. This forces a fresh OAuth flow and resolves most connection issues. If the problem persists, contact [support@sleekplan.com](mailto:support@sleekplan.com).