Connect your coding agent to BugBoard
One MCP server, 75 tools. Pick your editor, copy the config, and your agent can read and update your board.
Create a key
In BugBoard, open the MCP page and create a key. It starts with
bb_and is shown once. Save it asBUGBOARD_MCP_TOKENin your shell, not in the repo.Add the server
Run in your project
claude mcp add --transport http bugboard https://api.bugboard.ai/mcp/sse \ --header "Authorization: Bearer $BUGBOARD_MCP_TOKEN"Save as .mcp.json
{ "mcpServers": { "bugboard": { "type": "http", "url": "https://api.bugboard.ai/mcp/sse", "headers": { "Authorization": "Bearer ${BUGBOARD_MCP_TOKEN}" } } } }The command saves the server for you only. To share it with your team, commit the .mcp.json file: Claude Code asks each person to approve a project server the first time. In .mcp.json the type must be http, because an entry with a url and no type is skipped.
{ "mcpServers": { "bugboard": { "url": "https://api.bugboard.ai/mcp/sse", "headers": { "Authorization": "Bearer ${env:BUGBOARD_MCP_TOKEN}" } } } }.cursor/mcp.json inside a project applies to that project. Save it as ~/.cursor/mcp.json to have the server in every project. Cursor fills in ${env:NAME} in the headers from your environment.
{ "inputs": [ { "type": "promptString", "id": "bugboard-token", "description": "BugBoard MCP token", "password": true } ], "servers": { "bugboard": { "type": "http", "url": "https://api.bugboard.ai/mcp/sse", "headers": { "Authorization": "Bearer ${input:bugboard-token}" } } } }Portable format: save as .mcp.json
{ "mcpServers": { "bugboard": { "type": "http", "url": "https://api.bugboard.ai/mcp/sse", "headers": { "Authorization": "Bearer ${BUGBOARD_MCP_TOKEN}" } } } }VS Code asks for the key the first time the server starts, so the key is not in the file. If your setup reads the portable .mcp.json in the workspace root instead, use the second file: it has no prompt and reads the key from the environment variable. VS Code's reference says new Copilot configurations should use the destinations MCP: Add Server offers, and describes this servers format for existing setups.
{ "mcpServers": { "bugboard": { "serverUrl": "https://api.bugboard.ai/mcp/sse", "headers": { "Authorization": "Bearer ${env:BUGBOARD_MCP_TOKEN}" } } } }Windsurf's documentation names two locations: the file that Open MCP config file opens in the Cascade menu (~/.config/devin/mcp_config.json on macOS and Linux), and ~/.codeium/windsurf/mcp_config.json, which the editor uses for discovery. Edit the file that menu action opens. Windsurf fills in ${env:NAME} in the headers, and accepts url in place of serverUrl.
[mcp_servers.bugboard] url = "https://api.bugboard.ai/mcp/sse" bearer_token_env_var = "BUGBOARD_MCP_TOKEN"Codex reads MCP servers from the [mcp_servers] tables of ~/.codex/config.toml, or of .codex/config.toml in a project you trust. bearer_token_env_var names the variable whose value Codex sends as the bearer token.
Ask your agent
Try "list my open bugs in BugBoard". If it answers with your tasks, you are connected.
What your agent can do
75 tools, grouped. These are the names your agent sees.
Read the board
- list_projects
- list_tasks
- search_tasks
- get_task
- list_subtasks
- list_comments
- list_task_links
- list_project_columns
- list_project_tags
- list_org_members
File and update work
- create_task
- create_tasks
- draft_task
- update_task
- update_tasks
- delete_task
- change_task_type
- set_task_parent
- comment_task
- batch_add_comments
- edit_comment
- delete_comment
- link_tasks
- unlink_tasks
- subscribe
Avoid duplicates
- Finds tasks of the same type with a matching title, and close issues in the project's linked GitHub repository, before your agent files a new one.
Call it before create_task. search_tasks looks at titles and descriptions across all projects the key can see.
Plan with initiatives
- create_initiative
- list_initiatives
- link_task_to_initiative
- link_tasks_to_initiative
- unlink_task_from_initiative
- list_initiative_tasks
Tags, columns and notifications
- create_tag
- add_tag_to_task
- remove_tag_from_task
- add_tag_to_tasks
- remove_tag_from_tasks
- set_project_columns
- list_my_notifications
- mark_notifications_read
Track time
- start_timer
- stop_timer
- log_time
- get_running_timer
- list_time_entries
- update_time_entry
- delete_time_entry
- get_task_time
Manual testing
- list_test_areas
- add_test_areas
- create_test_area
- update_test_area
- delete_test_area
- create_test_feature
- update_test_feature
- delete_test_feature
- record_test_result
- link_task_to_test_feature
- unlink_task_from_test_feature
- list_test_features_for_task
Brainstorm canvas
- list_canvas_cards
- create_canvas_card
- update_canvas_card
- move_canvas_card
- delete_canvas_card
- vote_canvas_card
- promote_canvas_card
- comment_canvas_card
- assign_canvas_card
- add_tag_to_canvas_card
- remove_tag_from_canvas_card
- add_person_to_canvas_card
- remove_person_from_canvas_card
- link_task_to_canvas_card
- unlink_task_from_canvas_card
Check for duplicates before filing
Have your agent call check_duplicates right before create_task. It looks through the project's tasks of the same type and, when the project is linked to a GitHub repository, that repository's open and recently closed issues.
- Each match says where it comes from (bugboard or github), whether it is open or closed, links to it and has a score from 0 to 1. Both sources are scored the same way, so their scores compare.
- No linked repository? It checks BugBoard only.
- GitHub slow or rate limiting? You still get the BugBoard matches, with a note naming the repository it skipped.
- Issues are read through your organization's GitHub App, only for people who can see the project, and kept in memory for a few minutes. Nothing is stored.
Found a likely duplicate? Comment on that task with comment_task instead of filing a new one.
{
"name": "check_duplicates",
"arguments": {
"project_id": 7,
"type": "bug",
"title": "Checkout button does nothing on Safari"
}
}{
"count": 2,
"duplicates": [
{
"source": "github",
"repo": "acme/web",
"number": 412,
"title": "Safari: checkout button not clickable",
"state": "closed",
"url": "https://github.com/acme/web/issues/412",
"score": 0.71,
"label": "likely_duplicate"
},
{
"source": "bugboard",
"taskId": 41,
"type": "bug",
"title": "Checkout does nothing on Safari",
"status": "pending",
"state": "open",
"url": "https://bugboard.ai/projects/7/tasks/41",
"score": 0.64,
"label": "likely_duplicate"
}
],
"github": [
{
"repo": "acme/web",
"compared": 214
}
]
}You decide what each key can touch
A key can never do more than the person who made it. Narrow it further when you create it.
- Limit a key to read only, or to chosen projects, when you create it through the API. Keys made on the MCP page are not limited today.
- Projects outside the key do not show up at all.
- The key is shown once and stored only as a hash. Lost it? Re-issue it.
- Every call is checked, rate-limited and written to the audit log.