MCP resources
Reusable context for AI-assisted implementation work.
MCP resources are a section on a backlog item where AI tools can store and reuse context as work progresses, so that future work (by a person or another agent) can pick up with the right context.
They’re designed to capture AI-assisted work during implementation, including research, decisions, and changes made while building or fixing the item.
MCP resources are created and updated through the Atono MCP server, but can also be edited manually in the Atono web application. If you’re not using the MCP server tools while working on an item, MCP resources may not be available on that backlog item yet.
To use MCP resources from your IDE, your AI coding assistant must be connected to the Atono MCP server.
Access MCP resources
In the Atono web application, MCP resources (if available) are accessible from the details panel of a backlog item.
- Open a backlog item (story or bug).
- In the details panel, select MCP resources.
- In the MCP resources dialog, click the Investigation or Change summary tab.
What’s included
Investigation
Investigation is where research and context is captured while you’re still figuring out the work. Use it for a plan-first workflow, where agents (and developers) document what they learn before making any code changes. For example:
- Findings and constraints
- Relevant files, endpoints, or dependencies
- Open questions and decisions
- Recommended next steps
Change summary
Change summary is a running record of the changes made while completing the work. Use it once implementation begins, and keep it up to date as the implementation evolves. For example:
- What was added, changed, or removed
- Test updates
- Behavior changes
- Outcomes (what works now)
Using MCP resources from your IDE
MCP resources are most useful when your IDE assistant treats them as a shared “memory” for a backlog item.
Instead of relying on the chat history in your AI tool (which can get long, fragmented, or disappear between sessions), MCP resources store the important context directly on the story or bug. They're especially helpful when work spans multiple sessions, tools, or people.
You don’t need to call MCP tools directly. In most IDEs, you can tell your assistant what to fetch and what to update, and it will use the MCP server tools behind the scenes.
When prompting your IDE assistant, include the backlog Item ID (for example
STORY-98orBUG-123) so it knows which item to update.
Start by pulling context
When you begin working on a backlog item, pull the existing MCP resources first. This helps you avoid repeating research or redoing work that’s already been done. For example:
"Fetch the MCP resources for STORY-98 (both INVESTIGATION and SUMMARY) and summarize what I should know before I start coding."
Keep MCP resources updated while working
At the start of the session, tell your assistant to keep MCP resources current as it works. This is what turns MCP resources into a reliable handoff for the next person (or the next AI session). For example:
“While you work on STORY-98, keep MCP resources up to date as you go. Use INVESTIGATION for research, findings, constraints, and open questions. Use SUMMARY for implementation changes and test updates. Update MCP resources after each meaningful step, not just at the end.”
Capture investigation notes
Use this when the assistant discovers something you don’t want to lose—like a constraint, a dependency, or the real reason something is failing. For example:
“Update MCP resources for STORY-98. Add the key findings and constraints we discovered to INVESTIGATION.”
Capture implementation changes
Use this after a meaningful implementation step (new behavior, refactor, tests, schema changes). This keeps the backlog item’s Change summary accurate even if the work spans multiple sessions. For example:
“Update MCP resources for STORY-98. In SUMMARY, record the changes you just made, including files touched, behavior changes, and tests updated.”
Wrap up with a clean handoff
Before you stop working, do a quick cleanup pass so the next developer or agent can pick up the work without guessing what’s current. For example:
“Before we finish work on STORY-98, update MCP resources so they’re ready for handoff. In INVESTIGATION, remove anything outdated and list remaining open questions and next steps. In SUMMARY, make sure it reflects the current code by removing reverted changes, and include the current status.”
Using MCP resources in Atono
You can view and edit MCP resources directly in the Atono web application.
View MCP resources
- Open a backlog item (story or bug).
- In the right-side panel, click MCP resources.
- Select Investigation or Change summary.
Edit MCP resources
- Open the MCP resource you want to update (Investigation or Change summary)
- Modify the content.
- Click Save.
Version history
Atono tracks changes to MCP resources so you can review what changed over time and restore earlier versions if needed. In MCP resource history, the diamond (✦) icon indicates a version was created by an MCP tool.
MCP resources have their own version history, separate from a backlog item’s content version history. Changes to one do not affect the other.
Investigation and Change summary each have their own version history. Restoring or deleting versions in one tab does not affect the other.
Restore a version
- Open a backlog item (story or bug).
- In the details panel, select MCP resources.
- Select the Investigation or Change summary tab.
- In the version list on the right, hover over the version you want to restore.
- Select the ellipsis (...) icon and choose Restore this version.
- In the Restore version dialog, confirm by clicking Restore version, or click X to cancel.
Restoring creates a new version identical to the one you selected, and makes it the current version for that tab.
Delete a version
- Open a backlog item (story or bug).
- In the details panel, select MCP resources.
- Select the Investigation or Change summary tab.
- In the version list on the right, hover over the version you want to delete.
- Select the ellipsis (...) icon and choose Delete this version.
- In the Delete version dialog, confirm by clicking Delete version, or click X to cancel.
Deleting a version only affects the version history for the selected tab.
Updated 21 days ago
