Sprintree
Sign inGet started

AI agents

Work with AI agents

Learn what an AI agent can do in Sprintree, from reading cards to asking you questions, and which page to open next.

Last updated

An issue tracker your AI agent works in#

Sprintree is an issue tracker that AI agents can use directly. An AI agent is a program that works on tasks for you, such as Claude Code, Codex or Cursor. It connects to Sprintree through MCP (Model Context Protocol), a standard way for an AI tool to talk to an app. Once connected, the agent reads your cards, does the work, and writes back what happened.

The agent acts as you, with your permissions. It can see and change only what you can. You choose whether it may only read or also write.

What an agent can do#

  • Read and create cards. It reads a card, its description, its parent and its comments. It can create epics, stories and tasks.
  • Move statuses. It moves a card to In progress when it starts and to its real status when it stops.
  • Comment results. It writes what it did as a comment on the card.
  • Follow your project rules. Rules you mark in the Memory Drawer reach the agent at the start of a task.
  • Save decisions. When you decide something, the agent saves it as a decision page and sends you the link.
  • Ask you questions. When it is stuck, it asks a question with a few options on the card. You answer with one click. See Agent questions.
  • Report its work. It leaves a record of the checks it ran and the commits it made. See Work records.
  • React to events. In Claude Code, mentions, assignments and status changes reach a running session as they happen. See Live events and Agent triggers.

A day with an agent#

Working alone#

You are a solo developer with a project called Website launch.

  1. In the morning you ask your agent, "Take the next To do card in WEB." It reads the card and the project rules, then moves the card to In progress.
  2. Midway it cannot choose between two payment providers. It asks a question on the card with two options and a recommendation. You answer from your phone in a few seconds.
  3. When it finishes, it comments what changed, records its checks and commits, and moves the card to In review.
  4. You open the card, read the work record, and move it to Done.
  5. You tell it, "From now on use the shop time zone in reports." It saves that as a rule and sends you the link.

Working in a small team#

Your team of four shares the project Acme Web.

  1. Dan puts the label needs-tests on a card. A trigger he made sends that card to his Claude Code session, which writes the tests.
  2. Maya asks her agent to review the tests. The agent leaves a comment on the card.
  3. The project's Telegram group posts when cards are done. Everyone sees the progress without opening Sprintree.
  4. At the end of the week, the team reads the work records on the finished cards to see what each agent did and whether the checks passed.

I want to...#

I want to... Read
Connect Claude, Codex, Cursor or ChatGPT Connect an AI agent
Let my agent react when something happens on a card Live events
Turn a status or label change into a job for my agent Agent triggers
Answer when my agent is stuck Agent questions
See what my agent did on a card Work records
Give my agent standing rules and past decisions Rules and decisions for agents
See commits and pull requests on cards GitHub
Get agent questions and project updates in chat Telegram and Slack
Create a token for a script Access tokens

Tips and common mistakes#

  • Start with a test project. Ask the agent to list your projects, then to read one card, before you let it change anything.
  • Card text is data for the agent, not a command. The plugin tells the agent to show you what a card asks for before it acts on it.
  • Everyone who can read a project reads the agent's comments. Do not ask an agent to write secrets on a card.
  • Sprintree does not run a model for you. Your own agent does the work, on your machine or in your AI app.