bbkt (Bitbucket CLI & MCP Server)

A complete command-line interface and Model Context Protocol (MCP) server written in Go that provides programmatic integration with Bitbucket workspaces and repositories.

Features

Dual Mode: Run as a rich, interactive CLI tool for daily developer tasks, or as an MCP server for AI agents.
Git Awareness: Automatically detects your current Bitbucket repository from .git/config when run from the terminal.
Interactive UI: Sleek terminal UI wizards trigger automatically when required arguments are omitted.
Read/Write Operations: Seamlessly manage repositories, workspaces, pipelines, issues, and pull requests. Modify or delete repository source code directly from the API.
Authentication: Supports standard App Passwords or an interactive OAuth 2.0 web flow for desktop users.

Installation

From Source

# Clone the repository
git clone https://github.com/zach-snell/bbkt.git
cd bbkt

# Run the install script (builds and moves to ~/.local/bin)
./install.sh

Ensure ~/.local/bin is added to your system $PATH for the executable to be universally available.

From GitHub Releases

Download the appropriate binary for your system (Linux, macOS, Windows) from the Releases page.

CLI Usage

bbkt provides a robust command-line interface with the following core modules:

# Manage workspaces
bbkt workspaces [list, get]

# Manage repositories
bbkt repos [list, get, create, delete]

# Manage pull requests and comments
bbkt prs [list, get, create, merge, approve, decline]
bbkt prs comments [list, add, resolve]

# Trigger and view pipelines
bbkt pipelines [list, get, trigger, stop, logs]

# Issue tracking
bbkt issues [list, get, create, update]

# Read, search, and edit source code
bbkt source [read, tree, search, history, write, delete]

MCP Usage

The tool also serves as an MCP server. It supports two protocols: Stdio (default via bbkt mcp) and the official Streamable Transport API over HTTP.

Stdio Transport (Default)

If you intend to use this with an MCP client (such as Claude Desktop or Cursor), add it to your client's configuration file as a local command:

{
  "mcpServers": {
    "bitbucket": {
      "command": "/absolute/path/to/bbkt",
      "args": ["mcp"],
      "env": {
        "BITBUCKET_USERNAME": "your-username",
        "BITBUCKET_API_TOKEN": "your-api-token"
      }
    }
  }
}

Streamable Transport (HTTP)

You can run the server as a long-lived HTTP process serving the Streamable Transport API (which uses Server-Sent Events underneath). This is useful for remote network clients.

bbkt mcp --port 8080

Environment Variables

| Variable | Description | Required | |----------|-------------|----------| | BITBUCKET_USERNAME | Your Bitbucket username | No (but recommended for API Tokens) | | BITBUCKET_API_TOKEN | An Atlassian API Token | No (If omitted, triggers OAuth 2.0 browser flow) | | BITBUCKET_CLIENT_ID | OAuth 2.0 Client ID | Only if using OAuth | | BITBUCKET_CLIENT_SECRET | OAuth 2.0 Client Secret | Only if using OAuth |

API Token Scopes & Security

read:workspace, read:account, read:user, read:repository:bitbucket, write:repository:bitbucket, read:pullrequest:bitbucket, write:pullrequest:bitbucket, read:pipeline:bitbucket, write:pipeline:bitbucket

Token Introspection: The bbkt mcp server dynamically evaluates your API token's granted scopes at startup. If you omit specific permissions (like write:pipeline:bitbucket), the server will completely hide the associated MCP tools (trigger_pipeline, stop_pipeline) from the AI agent to prevent hallucinated successes.

Explicit Tool Denial: Even if your token has full admin privileges, you can explicitly deny the AI agent access to any tool using the BITBUCKET_DISABLED_TOOLS environment variable.

export BITBUCKET_DISABLED_TOOLS="delete_repository,delete_branch,delete_file"

Tools Provided

manage_workspaces: Getting and listing Bitbucket workspaces
manage_repositories: Listing, getting, creating, and deleting repositories
manage_refs: Listing, creating, and deleting branches and tags
manage_commits: Listing and getting commits, diffs, and diffstats
manage_source: Source code operations (read, list_directory, get_history, search, write, delete)
manage_pull_requests: All pull request operations (list, get, create, update, merge, approve, unapprove, decline, diff, diffstat, commits)
manage_pr_comments: Managing pull request comments (list, create, update, delete, resolve, unresolve)
manage_pipelines: Managing Bitbucket Pipelines (list, get, trigger, stop, list-steps, get-step-log)
manage_issues: Managing repository issues (list, get, create, update)

Development

Requirements:

Go 1.25+

# Run tests
go test ./...

# Run the linter
golangci-lint run ./...

License

This project is licensed under the Apache 2.0 License.

MCP Servers