Standup MCP Server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| STANDUP_URL | Yes | The Standup server URL, for example https://thestandup.app. | |
| STANDUP_API_KEY | Yes | Your Standup API key. It starts with sk- and is created from the Standup API Keys page. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
You are connected to a Standup kanban board. Cards flow left-to-right through user-configured columns.
Running the pipeline
When asked to work on or process a board, work every workable card through the stages. After advancing a card, carry on with it in its new column as that column's agent(s).
Start each pass with
get_board_summary. Useget_boardorget_commentsonly for full content: card text, comment bodies.Before telling the user a board's or card's state, call
get_board_summaryagain.
Board context
Before working, read the board's
context(project background, constraints) andrepository(git URL to clone if you are not in the repo) fromget_board_summary.Board context is optional: it adds project-specific rules, and the pipeline runs without it.
Pipeline stages
A column with at least one enabled agent is an active stage. Run each column agent as its own subagent, following that agent's
instructions.A column with no enabled agent belongs to the user, wherever it sits: only process / move cards where instructed by the user or agent instruction.
Assignment
Each column agent is its own assignee:
agent:<identifier>, using the agent'sidentifier. Never assign agents under your own username.Agents in the same column work on a card in parallel: agent assignments never stop you, and a card assigned to one of your agents is yours to continue.
When looking for work, take cards in this order:
Cards assigned to one of your agents where a person has replied since that agent's last comment (handed back).
Unfinished cards in your agents' columns that no person is assigned to and that agent has not finished on this visit.
Assign yourself before you start, even if you only approve. When you finish, comment with your
agentId, then unassign yourself.Judge from the card's comments and their times whether you have finished. A rejected card can return, so an approval older than later work no longer counts. If unclear, ask the user via comments on the card.
Advancing and rejecting
Reject: move the card to the previous column with a comment (with your
agentId) explaining why.Advance: only when every enabled agent in the column has finished and no agent is assigned. The last agent to finish moves it.
Where the work lives
When a card needs code changes, unless the board context says otherwise: work in a local clone of
repository, one branch per card named for the change in conventional commit style (e.g.fix/files-list,feat/deep-links), commit, and do not push or open pull requests.Whenever you commit, push or open a PR, call
set_card_workwith the absolute repository path, branch and new commit SHA(s), so the card shows where its work is.Comments can then stay short, e.g. "Committed
811fca4onfix/files-list".
Blockers
Comment with your
agentId,optionsfor clickable choices and an@mentionof the user's Standup username (the API lists valid ones). Assign the user, unassign yourself, and leave the card in its column.Ask only on cards, never in the chat: a question there stops the board being processed.
Waiting on people
Never work on a card assigned to a person.
When the person answers, Standup hands the card back: it unassigns them and reassigns the agent that asked. Check you have enough info before continuing.
If a person has not answered a card for a long time, comment and
@mentionthem.When a tool response carries a "Handed back to you" note, finish your current step, then resume those cards before new work. Between cards, check for handed-back cards.
If the user asks why a handed-back card was not picked up: agents only act while a session runs, and Standup cannot start one. Suggest keeping a session polling the board, e.g. in Claude Code
/loop 15m process the <board name> board, or a cron job running that prompt.
Triage
For a triage stage: advance well-defined cards; ask about ambiguous ones as a blocker; comment on empty or nonsensical ones without advancing.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_boardsB | List all boards with name and column summary |
| create_boardA | Create a board you own from a template: blank (Todo, In Progress, Done) or ai-software-development (Inbox, Triage, In Progress, Code Review, Testing, Done, with the default agents). Returns the new board with its column and agent identifiers. |
| get_boardB | Get a full board with columns, cards, context, and agents |
| get_board_summaryA | Get a compact board: name, context, repository, columns with agents, and one row per card in every column with its first line, assignees, stickers, work, comment count and last comment author, agentId and time. No comment bodies: use get_comments or get_board for full content. |
| create_cardC | Create a new card in a column |
| update_cardB | Update card text. Automatically finds the column containing the card. |
| delete_cardB | Delete a card. Automatically finds the column containing the card. |
| move_cardB | Move a card to a different column. Automatically finds the source column. |
| assign_cardA | Assign a user or agent to a card. Adds to existing assignees without duplicates. Column agents are assigned as agent:. |
| unassign_cardB | Remove a user or agent from a card. |
| search_cardsB | Search for cards across all boards by text content |
| get_commentsB | Get all comments on a card. A deleted comment keeps its place, author and time, with empty text and a deletedAt time. |
| add_commentB | Add a comment to a card, with optional @mentions and clickable option buttons |
| update_board_contextA | Set a board's project context (markdown) and/or repository URL. Omitted fields are left unchanged. |
| update_column_agentsA | Set the agents for a column. The list replaces the column's existing agents, so to change one agent send the full list from get_board with that agent edited. Keep each existing agent's identifier so its assignments and comments stay linked. |
| set_card_workA | Record where the work for a card lives: repository path, branch, commits, whether it was pushed and the pull request link. Merges with what is already recorded: new commits are added and fields you leave out are kept. |
| add_columnB | Add an empty column to a board |
| rename_columnC | Rename a column |
| reorder_columnB | Move a column to a new position on its board, keeping its cards |
| delete_columnA | Delete a column. Refused while it has cards, unless deleteCards is true, which permanently deletes them too. |
| add_stickerB | Add a sticker (an emoji) to a card. Automatically finds the column containing the card. |
| remove_stickerB | Remove a sticker from a card. If the card has the same sticker more than once, one is removed. |
| get_attachmentA | Get a file attached to a card, using the fileId from the card's attachments in get_board. Images are returned so you can see them. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 23 tools
Each tool targets a clearly distinct resource/action across boards, columns, cards, comments, stickers, attachments, assignments, and work metadata. Overlapping-looking tools like get_board vs get_board_summary and assign_card vs update_column_agents are clearly differentiated by target scope.
All tools follow a consistent snake_case verb_noun pattern, including compound nouns such as get_board_summary, update_board_context, update_column_agents, and set_card_work. There is no mixing of camelCase or vague standalone verbs.
23 tools is on the high side, but the domain is broad and most operations correspond to a real board-management need. A few retrieval or sticker/comment operations could potentially be consolidated, but the set remains reasonably scoped rather than bloated.
Core card and column lifecycle operations are well covered, including move, assign, reorder, and work tracking. However, board deletion/rename, comment editing/deletion, and attachment creation/deletion are missing, and get_attachment implies attachments can exist without a corresponding tool to add them.