mait.sh1.69.2

Private beta: features and content may change. No warranty; use at your own risk.

Docs / Agent tools (MCP)

MCP AIDE

Let a compatible agent inspect AIDE projects, operate local clusters and change development routes.

For teams letting agents prepare or diagnose local Kubernetes development environments.

Read from the top for the basics; the later sections go deeper.

Beta
All components

What it doesBasics

  • Project and runtime discovery
  • Cluster status and lifecycle
  • Service route inspection
  • External and internal route switching

First useBasics

  1. Inspect one project's cluster and route state before changing it.
  2. Inspect cluster and route status.
  3. Apply the smallest lifecycle or routing change needed.

Everyday tasksEveryday use

Route a service to a live development process

A developer needs fast reloads while the rest of the application remains in the local cluster.

  1. Query project, cluster, and current route state.
  2. Confirm the local process is reachable from the development environment.
  3. Apply an external route for the intended service and namespace.
  4. Query route state and exercise the real ingress URL.
  5. Reset the override when the live-editing session ends.

What you getIngress traffic reaches the live process temporarily without changing the application ingress definition.

How to check: Before-and-after route status plus a request through the project URL.

Delete a disposable cluster with separate storage intent

A finished development environment should be removed without accidentally purging persistent volumes.

  1. Query cluster state and identify the exact project context.
  2. Request deletion without confirmation and read the confirmation message.
  3. Re-query the same project and cluster, then repeat with delete confirmation.
  4. Add purge and its separate confirmation only when volume removal is also approved.
  5. Query the cluster list to verify cleanup.

What you getCluster and storage cleanup follow separately reviewed decisions.

How to check: Same-context project and cluster query, confirmation message, explicit confirmation fields, command result, and final cluster inventory.

Settings to decideAdvanced

AIDE executable
Use the default executable only when AIDE is reliably available on the server PATH; otherwise set its reviewed launch path.
Project selection
Pass projectDir, projectFile, or both so every operation resolves the intended AIDE project rather than ambient server state.
Route namespace
Choose the namespace containing the service, which can differ from the default apps namespace.
Cleanup confirmation
Treat cluster deletion, volume purge, and purge confirmation as distinct decisions.

CommandsAdvanced

$ aide_query({"action":"setup","projectDir":"/workspace/project"})
$ aide_query({"action":"doctor","projectDir":"/workspace/project"})
$ aide_cluster({"action":"status","projectDir":"/workspace/project"})
$ aide_route({"action":"status","service":"web","projectDir":"/workspace/project"})

Problems and fixesHelp

What you seeCheckFix
Discovery or lifecycle operations cannot find the expected project.Inspect projectDir and projectFile and confirm the file is valid in that context.Correct the request-scoped project context and rerun readiness discovery before mutation.
A route is set but traffic does not reach the live process.Inspect route status, namespace, target host and port, and process reachability from the cluster boundary.Correct the endpoint or restore the internal route; verify through the project ingress rather than localhost.
Cluster deletion succeeds or previews but volume purge is refused.Confirm that purge and confirmPurge express a separately approved storage decision.Review the persistent-data impact and repeat only with both purge fields when removal is intended.
What you see
Discovery or lifecycle operations cannot find the expected project.
Check
Inspect projectDir and projectFile and confirm the file is valid in that context.
Fix
Correct the request-scoped project context and rerun readiness discovery before mutation.
What you see
A route is set but traffic does not reach the live process.
Check
Inspect route status, namespace, target host and port, and process reachability from the cluster boundary.
Fix
Correct the endpoint or restore the internal route; verify through the project ingress rather than localhost.
What you see
Cluster deletion succeeds or previews but volume purge is refused.
Check
Confirm that purge and confirmPurge express a separately approved storage decision.
Fix
Review the persistent-data impact and repeat only with both purge fields when removal is intended.

All MCP AIDE problems and fixes →

Good to knowHelp

  • It controls local AIDE environments.
  • Privileged host setup stays with the AIDE command line.
  • Delete and purge each ask for their own confirmation; query the project and cluster first to see what they hold.