DEV Community

Karim GAAD
Karim GAAD

Posted on Originally published at flowboard.dev

FlowBoard MCP Server: Let AI Manage Your Projects

What Is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources. Instead of copy-pasting task IDs and switching between your IDE and FlowBoard, an MCP server lets your AI tool talk to FlowBoard directly.

Ask Claude to "create a bug for the login timeout issue" and it calls the FlowBoard API for you. Ask it to "show me all open bugs assigned to me" and it queries your board instantly. No browser tab required.

What You Can Do

The FlowBoard MCP server exposes 27 tools across 6 categories that any MCP-compatible AI can use:

The FlowBoard Ideas Board an MCP-connected assistant can read and act on, with Accept, Rework and Reject as the available decisions

Tasks & Projects

Tool What It Does
list_projects List all projects in your workspace with IDs, names, and colors
search_tasks Search and filter tasks by board (flows/bugs/ideas), status, project, assignee, type, tags. Use mine: true to list your own tasks. Supports comma-separated statuses and types for OR logic.
get_task Get a single task by ID or task number, including full details and all comments
create_task Create a new task. Ideas and bugs go to their respective boards. Flow types (feature, improvement, fix, chore) require a projectId.
update_task Change status, add a comment, link a GitHub PR or GitLab MR, assign someone, or set a branch name, all in a single call
delete_task Soft-delete (trash) a task. Can be restored from the UI.
bulk_update_tasks Apply one change to up to 50 tasks at once: status, assignee, project, tags, due date, plus an optional comment on each
list_members List workspace members and roles, so an agent can resolve a name to the user ID that assignment expects
get_workspace_schema Read the priority formula, configured custom fields, projects, and the ideas and bugs board IDs, for discovery before writing
score_task Set impact, clients, tech ease, estimate or custom fields and recompute the priority score. Leaves a before/after audit comment.

Idea Review

These four require an API key with the Review Ideas scope, and turn the server into something that can run a product owner's triage session. See letting an AI product owner triage your backlog.

Tool What It Does
accept_idea Accept an idea, moving it to ready
reject_idea Reject an idea. A reason is required and is posted as a comment the submitter can read.
rework_idea Send an idea back for more detail, also with a required reason
place_idea Move an accepted idea onto a specific flow project

Ideas & Comments

Tool What It Does
list_ideas List ideas with optional status filter
vote_idea Toggle a vote on an idea. Voting again removes the vote.
get_comments Get comments for a task (up to 100)
add_comment Add a comment to a task. Attributed to your account when FLOWBOARD_USER_EMAIL is set.

Releases & Release Drafts

Tool What It Does
list_releases List confirmed releases with version, name, task count, and dates
get_release Get a single release with all its associated tasks
add_tasks_to_release Attach tasks to a release that already exists, by release ID or version
remove_tasks_from_release Detach tasks from an existing release
get_release_draft View the current draft: tasks staged for the next release and hotfix, version, and deploy notes
add_to_release_draft Add tasks to the release or hotfix draft by ID or task number. Tasks move between lists automatically.
remove_from_release_draft Remove tasks from the release or hotfix draft
set_release_version Set the version string (e.g. "2.1.0") for the draft
confirm_release Publish a release from the draft. Archives all included tasks and clears the draft. Optionally set environments (Staging, Production) and public visibility.

Prompts and Resources

Beyond tools, the server ships two prompts. review_idea_backlog for walking the whole open ideas queue and triage_idea for a single idea by number, so you don't have to describe the workflow yourself. Both read a review rubric exposed as a resource at flowboard://review-rubric. A default ships with the package; point FLOWBOARD_REVIEW_RUBRIC at your own Markdown file to replace it with your product's actual criteria. The project list is likewise available at flowboard://projects.

All task-related tools accept either a taskId (Firestore document ID) or a taskNumber (the human-readable integer, e.g. 42). Task numbers are easier to work with, since agents can extract them from git branch names like feature/42-fix-login.

Setup (5 Minutes)

1. Generate an API Key

Go to Settings → Integrations → API Keys in your FlowBoard workspace. Click "Generate API Key", give it a name, and enable the permissions you need. Copy the key immediately - it's only shown once.

2. Configure Your AI Tool

Add this to your AI tool's MCP server configuration:

macOS / Linux:

{ "mcpServers": { "flowboard": { "command": "npx", "args": ["-y", "@flowboardlabs/mcp-server"], "env": { "FLOWBOARD_API_KEY": "your-api-key-here", "FLOWBOARD_USER_EMAIL": "you@example.com" } } } } Windows:

{ "mcpServers": { "flowboard": { "command": "cmd", "args": ["/c", "npx", "-y", "@flowboardlabs/mcp-server"], "env": { "FLOWBOARD_API_KEY": "your-api-key-here", "FLOWBOARD_USER_EMAIL": "you@example.com" } } } } FLOWBOARD_USER_EMAIL is optional but recommended. It lets the AI know who you are, enabling features like "show me my tasks" and attributing comments to your account instead of a generic "api" author.

Claude Code: Add to .mcp.json in your project root Cursor: Add to your Cursor MCP settings Other tools: Check your tool's MCP configuration docs

3. Start Using It

Once configured, your AI tool automatically discovers the FlowBoard tools. Just ask naturally:

  • "Show me my in-progress tasks"
  • "Show me all open bugs in the backend project"
  • "Create a bug: login page returns 500 when email contains a plus sign"
  • "Move task 42 to in_progress and assign it to me"
  • "What are the latest releases?"
  • "Add a comment to task 15: Fixed in PR #42"
  • "Add tasks 10, 12, and 15 to the release draft"
  • "Set version to 2.1.0 and confirm the release"
  • "Show me what's in the release draft"

How It Works Under the Hood

The MCP server is a lightweight Node.js process that communicates over stdio. It translates MCP tool calls into HTTP requests against FlowBoard's Cloud Functions API - the same endpoints you'd call with curl or any HTTP client.

Architecture:

AI Tool (Claude/Cursor) ↕ MCP Protocol (stdio) FlowBoard MCP Server ↕ HTTPS FlowBoard Cloud Functions API ↕ Firestore Database The server never accesses your database directly. All data flows through the authenticated API endpoints with the same rate limiting (100 req/min) and permission checks as any other API consumer.

Permissions and Security

API keys have granular permissions. When generating a key for MCP, enable the permissions your workflow needs:

  • Submit Ideas / Submit Bugs: create new tasks
  • Search Tasks: query and read tasks, comments, releases
  • Update Tasks: change status, assign, delete, vote
  • View Projects: list projects in the workspace
  • Manage Comments: add comments to tasks
  • Manage Releases: view/edit release drafts, add tasks to drafts, confirm releases

For a full-featured MCP agent, enable all permissions. For a read-only setup (e.g., for a reporting tool), enable only Search Tasks and View Projects.

Use Cases

Autonomous Bug Fixing

Combine the MCP server with Claude Code's agentic capabilities. The AI reads bugs from FlowBoard, writes fixes, creates PRs, and updates task status - all without you switching context. See our detailed guide on AI agent bug fixing for the full workflow.

Standup Prep

Ask your AI: "What did I work on yesterday and what's next?" It queries FlowBoard for your recent tasks, checks statuses, and summarizes your standup in seconds.

Triage From Your Editor

Working on a feature and notice a bug? Tell Claude: "Create a bug: the date picker doesn't handle timezones correctly, tag it with 'frontend' and 'date-handling'". Done. No tab switching.

Release Management

After finishing a sprint, tell your AI: "Add all done tasks to the release draft, set version to 2.1.0, and confirm the release." The agent searches for done tasks, stages them, and publishes the release, all without leaving your editor. You can also ask it to summarize what changed or prepare hotfixes for urgent bugs.

Frequently Asked Questions

Which AI tools support MCP?

Claude Code and Cursor have native MCP support. Any tool that implements the Model Context Protocol can connect to the FlowBoard MCP server.

Does the MCP server need to run all the time?

No. Your AI tool starts the MCP server process when needed and stops it when idle. The npx command handles installation and execution automatically.

Is my data sent to third parties?

The MCP server runs locally on your machine. It communicates with FlowBoard's API (your data) and with your AI tool (which you've already authorized). No additional third parties are involved.

Can I use this with self-hosted FlowBoard?

Yes. Set the FLOWBOARD_API_URL environment variable to point to your instance's Cloud Functions URL.

Get Started

Generate an API key in your workspace settings, paste the MCP config into your AI tool, and start managing your FlowBoard projects with natural language. The whole setup takes about five minutes.

Top comments (0)