Expose schedule manual-run result and schedule deployment log tools

[agentHits]

Summary

Dokploy MCP should expose enough schedule/deployment tooling for agents to safely run and inspect one-off compose schedule commands.

Related core issue: Dokploy/dokploy#4718.

Current behavior

The MCP exposes:

• schedule-create
• schedule-runManually
• deployment-allByType
• compose-readLogs, but it requires a Docker containerId

For a failed compose schedule manual run, MCP currently returns only:

{
  "success": false,
  "error": "Failed to execute schedule-runManually",
  "details": "Error: Request failed with status code 500"
}

A schedule deployment row can be queried via deployment-allByType({ type: "schedule", id: scheduleId }), and it contains deploymentId, status, and logPath, but MCP does not expose a direct deployment-readLogs tool in the current tool surface available to the client. compose-readLogs is not enough because it needs a runtime container id, not a schedule deployment id.

Needed MCP capabilities

Once Dokploy core returns structured metadata from schedule.runManually, MCP should expose it directly, for example:

schedule.runManually({ scheduleId }) => {
  scheduleId: string
  deploymentId: string
  status: "running" | "done" | "error"
  logPath?: string
}

MCP should also expose a safe log read tool for schedule deployments, either by generating the existing OpenAPI endpoint or adding a documented route:

deployment.readLogs({ deploymentId, tail? })

or a schedule-focused wrapper:

schedule.readLogs({ scheduleId, deploymentId?, tail? })

Why this matters

This is important for safe AI-agent automation:

• run a one-off diagnostic/smoke command inside a compose service;
• avoid persistent env mutation for temporary flags;
• avoid full env replacement when env is redacted;
• avoid SSH access to Dokploy internals;
• inspect failure logs without exposing secrets.

Example use case: run a dev-only acceptance smoke with process-local env flags inside a compose service, then verify logs and leave persistent env unchanged.

Acceptance criteria

• schedule-runManually returns or preserves actionable deployment metadata from Dokploy core.
• MCP exposes a way to read schedule deployment logs by deployment id.
• Failed schedule commands still expose log access metadata without returning raw secrets.
• Tests cover a failed manual run response and log retrieval routing.
• No fallback requires reading/replacing full env blocks.

Related existing env safety work

Safe env patch/upsert is tracked separately:

• Add safe partial environment variable upsert API and MCP tool dokploy#4525
• feat: add safe application env upsert dokploy#4581
• Add safe application env upsert MCP tool #55
• feat: add safe application env upsert tool #56

This schedule/log issue is complementary: it lets clients run temporary process-local commands without requiring persistent env mutation at all.