Skip to main content
Glama
fenbs-ai

fenbs-mcp

Official
README.md
# fenbs-mcp

Connect any MCP client to your [fenbs](https://fenbs.ai) board.

fenbs is the simple project management tool where people and AI assistants are both members of a
board, each holding a role, and every change is signed by who made it. Four lanes, three kinds of
task (Feature, Enhancement, Bug), roles in plain words, and an audit trail that names the assistant.
Decisions with sign-off, testing recorded on each task, and tasks a person can pre-approve for an AI
assistant to pick up when it is free.

## You probably do not need this package

fenbs speaks MCP over HTTP with a browser sign-in. Most clients connect to it directly, with no
token to copy:

| Client | How |
|---|---|
| Claude Code | `claude mcp add --transport http fenbs https://fenbs.ai/api/mcp` then `/mcp` |
| Claude (desktop / web) | Settings → Connectors → add `https://fenbs.ai/api/mcp` |
| Cursor | `mcp.json`: `{"mcpServers":{"fenbs":{"url":"https://fenbs.ai/api/mcp"}}}` |
| VS Code (Copilot) | `.vscode/mcp.json`: `{"servers":{"fenbs":{"type":"http","url":"https://fenbs.ai/api/mcp"}}}` |
| ChatGPT | Settings → Connectors → custom connector with the URL, where your plan allows it |

Step-by-step pages: <https://fenbs.ai/integrations>

## When you do

Use `fenbs-mcp` when the client only speaks **stdio**, or the machine cannot open a browser (CI, a
server, a container). It bridges stdin/stdout to the remote server with a token you issue by hand.

1. In fenbs: Settings → **Connect an AI assistant** → name it, tick read / write / comment → copy the
   token. It is shown once.
2. Run:

```sh
FENBS_TOKEN=your-token npx -y fenbs-mcp
```

Or in a client config that wants a command:

```json
{
  "mcpServers": {
    "fenbs": {
      "command": "npx",
      "args": ["-y", "fenbs-mcp"],
      "env": { "FENBS_TOKEN": "your-token" }
    }
  }
}
```

## What the assistant can then do

Tasks: `fenbs_whoami` · `fenbs_list_boards` · `fenbs_list_items` · `fenbs_get_item` · `fenbs_create_item` ·
`fenbs_update_item` · `fenbs_flag_item` · `fenbs_comment` · `fenbs_search` · `fenbs_delete_item` · `fenbs_restore_item`

Pre-approved for AI: `fenbs_next_approved_task` · `fenbs_release_task`

Decisions and rules: `fenbs_list_decisions` · `fenbs_get_decision` · `fenbs_add_decision` · `fenbs_update_decision` ·
`fenbs_request_signoff` · `fenbs_delete_decision` · `fenbs_restore_decision`

AI context: `fenbs_get_context` · `fenbs_add_context_note` · `fenbs_list_context_notes` ·
`fenbs_update_context_note` · `fenbs_delete_context_note`

**Rules are decisions.** A rule is a decision that holds from now on (`standingRule: true`). When a
person says how something must always, or never, be done, the assistant records it with
`fenbs_add_decision` (status `decided`, the person as decider, their words quoted), never as a context
note. `fenbs_get_context` shows the rules first and in full; assistants follow every one. Context notes
are only for facts nobody chose: how the code works, a trap, where something lives.

Projects: `fenbs_list_projects` · `fenbs_add_project` · `fenbs_rename_project` · `fenbs_delete_project` ·
`fenbs_restore_project`

The list comes from fenbs itself, so a new tool arrives without updating this package.

It acts as the person who issued the token and holds their role on the board, narrowed by the
scopes they ticked. A refusal is a sentence it can repeat to you: *"You are a Viewer on Kitchen
Refit, which cannot move tasks between lanes."* Revoke the token in Settings and it stops at once.

The whole guide, written for people and assistants alike: <https://fenbs.ai/llms.txt>

## Options

```
FENBS_TOKEN      the token (or --token <token>)
FENBS_MCP_URL    the endpoint (or --url <url>); default https://fenbs.ai/api/mcp
--version, --help
```

No dependencies. Node 20 or newer. MIT.

TDQS

A3.5/5.0

Scored across 30 tools

Disambiguation4/5

Most tools have clearly distinct purposes, with detailed descriptions distinguishing decisions, context notes, tasks/items, projects, flags, and approvals. Minor ambiguity remains because the same task resource is called 'item' in tool names but 'task' in other tools, and rules can be read via both fenbs_get_context and fenbs_list_decisions.

Naming Consistency4/5

All tools use a consistent fenbs_ prefix and snake_case, and most follow a verb_noun pattern such as fenbs_list_items or fenbs_update_decision. The main deviation is terminology: task-oriented tools use 'task' while the core CRUD tools use 'item', plus fenbs_whoami, fenbs_comment, and fenbs_search break the verb_noun pattern slightly.

Tool Count2/5

With 30 tools, the surface is heavy for the apparent scope, exceeding the threshold where each tool clearly earns its place. Many delete/restore/update variants across items, decisions, context notes, and projects could be consolidated or better grouped.

Completeness4/5

The server covers CRUD and lifecycle operations well for tasks/items, decisions, context notes, and projects, including restore, flagging, commenting, sign-off requests, and pre-approved task flow. Minor gaps exist around board-level administration, file operations, and direct sign-off actions, but core agent workflows are supported.

Maintenance

ActivityMaintained
ResponsivenessNo issues