> ## Documentation Index
> Fetch the complete documentation index at: https://docs.deepmerge.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent Context Sprints

> Use an Agent Context Sprint to interview one workflow, turn hidden operating knowledge into durable agent context, and prove your tools can reuse it.

An **Agent Context Sprint** is a focused way to seed Deepmerge with the workflow
knowledge your agents need but your docs usually miss: edge cases, handoffs,
judgment calls, customer exceptions, system quirks, and the reasons behind them.

Instead of asking people to write documentation, Deepmerge runs short AI-guided
interviews, synthesizes the evidence, and turns the results into durable memories
your tools can reuse.

## What a sprint produces

Each sprint gives you:

* A report synthesized from the completed conversations.
* Findings, decisions, artifacts, insights, and draft skills saved to the workspace.
* A participant plan and hypotheses, so the report is clear about who was asked
  and what the sprint tested.
* A reuse proof task: a concrete agent task that shows whether Claude, ChatGPT,
  Cursor, or another connected tool can apply the new context.

## When to run one

Run a sprint when a workflow depends on knowledge trapped in people's heads:

* Support escalation rules.
* Refund and exception handling.
* Fulfillment edge cases.
* Agency client onboarding.
* QA handoffs.
* Any repeatable process where agent mistakes cost time or money.

## How it works

<Steps>
  <Step title="Pick one workflow">
    Choose the workflow where better agent context would save time or reduce errors.
  </Step>

  <Step title="Invite the people who know it">
    Add participants who have first-hand examples. Deepmerge sends personal interview links.
  </Step>

  <Step title="Let the AI interviewer gather evidence">
    Each participant completes a short text or voice conversation. The interviewer
    searches the workspace during the conversation so questions use what the team
    already knows.
  </Step>

  <Step title="Read the report and reuse the context">
    As conversations finish, the report is rewritten from the evidence so far. When
    the sprint is done, use the saved insights and skills in a real agent task.
  </Step>
</Steps>

Context Sprints are available from the Deepmerge dashboard. For the normal agent
habit outside a sprint, see the [daily workflow](/workflow).