Give your AI agent sprint tools it can call

An AI agent helping you plan a sprint can call SprintFlint’s free tools, calculate with your team’s data and explain the result. The CLI gives it a direct way to discover each tool, read its contract and run it from the terminal.

SprintFlint’s free tools now have a command-line interface for that workflow. An agent can discover the available tools, inspect their input and output schemas, submit JSON and read the result. The calculations, questionnaire rules and generator presets come from the same modules as the browser tools.

The CLI runs locally. After downloading it, calculations require no network connection, SprintFlint account, API token or AI API credits. Your agent can use its own model to decide which tool to call and explain the result; the CLI performs the calculation or generates the template.

Give the agent a callable tool

Download the standalone file and run it with Node 24:

curl --fail --location --output sprintflint-free-tools.mjs \
  https://sprintflint.com/downloads/sprintflint-free-tools.mjs

node sprintflint-free-tools.mjs discover
node sprintflint-free-tools.mjs describe velocity-calculator

The download contains only the free-tool code. It does not require access to SprintFlint’s private application repository or an npm installation.

discover returns a machine-readable catalogue. describe TOOL gives the agent a description, input schema, output schema, supported formats and an example. Questionnaire tools also expose their questions and allowed answers, so the agent can ask for missing information instead of inventing a response.

An agent with terminal access can call the executable as a subprocess. Successful calculations return JSON on stdout. Invalid input returns a structured JSON error on stderr and a non-zero exit code. A wrapper can inspect the error, correct an input or ask the user for the missing value before trying again.

Calculate with the team’s history

Suppose the user asks:

Use our recent completed sprint points to estimate velocity, then forecast how long 120 remaining points might take. Our sprints last two weeks and the next starts on 5 October 2026. Show the assumptions.

The agent first inspects the velocity tool, then sends the actual sprint history as JSON:

node sprintflint-free-tools.mjs run velocity-calculator --input - <<'JSON'
{
  "rows": [
    { "name": "Sprint 1", "committed": 24, "completed": 22 },
    { "name": "Sprint 2", "committed": 26, "completed": 24 }
  ]
}
JSON

That result includes an average of 23 completed points and a sample standard deviation of about 1.414 points. The agent can pass those returned values to the forecast tool:

node sprintflint-free-tools.mjs describe sprint-forecaster
node sprintflint-free-tools.mjs run sprint-forecaster --input - <<'JSON'
{
  "remainingPoints": 120,
  "velocity": 23,
  "velocityStdDev": 1.4142135623730951,
  "sprintLength": 2,
  "startDate": "2026-10-05"
}
JSON

Both dates in this small example land on 28 December, because the tool rounds each estimate up to whole sprints. The agent should explain that rounding and preserve the underlying results. Two sprints are a tiny history: this demonstrates how to connect the tools, rather than providing a strong forecast for a real team. The forecast uses a normal-approximation planning heuristic; its p50 and p90 labels do not guarantee delivery dates.

Chain tools without hiding the assumptions

The CLI covers all 15 free tools: velocity, capacity, cycle time, forecasting, story-point estimates, maturity assessment, framework selection, retrospective selection, goal generation and validation, story splitting, definitions of done, review agendas, burndowns and MCP configuration.

An agent can use these in a planning conversation:

• Read velocity from the team’s completed sprints.
• Ask about team availability and use the capacity calculator.
• Check a proposed sprint goal and show the criteria behind its score.
• Generate an editable definition of done or review agenda.
• Export a burndown SVG for the planning notes.

Units are part of the contract. Capacity uses working days and total team PTO person-days; the forecaster uses weeks. Story points belong to the team that estimates them. An agent should ask about missing availability or units, and keep the tool input alongside its result.

Generators produce templates and rules-based suggestions for a team to review. The MCP configuration generator leaves YOUR_TOKEN as a placeholder and never retrieves credentials.

Add it to your agent’s workflow

Give your terminal-enabled agent the downloaded file and this instruction:

Use SprintFlint’s free-tools CLI for sprint calculations. Start with discover, inspect the chosen tool with describe, and call run with JSON on stdin. Use my actual data, ask for missing inputs, and report the tool result and assumptions. Check the exit code before using a result.

This is a local executable your agent calls through its terminal tools. The standalone CLI does not register an MCP server or write SprintFlint tickets. For authenticated workspace operations, use SprintFlint’s separate MCP integration.

Use the agent guide for the command and error protocol, or try the same calculations in the free browser tools.