IT operations & ticketing

IT operations teams spend much of their time on repetitive diagnostics, routing, and low-risk changes. An Alquimia Platform agent can handle the first line of these requests while keeping sensitive actions behind approval gates.

What the use case looks like

An employee opens a ticket because they cannot access an application. The agent:

Reads the ticket from the ITSM tool.
Runs a diagnostic tool to check the user’s account status.
If the issue is a common lockout, performs the unlock after manager approval.
Updates the ticket with the action taken and closes the loop with the employee.

Registering tools with least-privilege policies

IT tools often bundle read and write operations on the same connection. In Alquimia Platform, each operation is classified separately in the registry. A freshly registered connection grants access to nothing until operations are explicitly classified.

# Register the identity-provider MCP connection
alquimia registry tools add identity-tools \
  --provider-id mcp \
  --connection-config '{"url": {"$secretRef": "IDP_MCP_URL"}}'

# Audit and classify operations one at a time
alquimia registry tools set-operation identity-tools check_user_status \
  --severity read-only --tier-grants reader,editor,operator

alquimia registry tools set-operation identity-tools unlock_user_account \
  --severity mutating-recoverable --tier-grants editor,operator --approval-required

alquimia registry tools set-operation identity-tools reset_user_password \
  --severity destructive --tier-grants operator

The classification table drives enforcement:

Severity	Who can run it	Approval
`read-only`	`reader`, `editor`, `operator`	optional
`mutating-recoverable`	`editor`, `operator`	optional
`destructive`	`operator` only	always required

The agent spec references the tool by ID and declares a role ceiling:

{
  "assistant_id": "itops-agent",
  "role": "editor",
  "response": {
    "provider_id": "alquimia",
    "profile": {
      "system_prompt": "You are an IT operations assistant. Read tickets, run diagnostics, and only perform approved changes.",
      "tools": [
        {
          "provider_id": "mcp",
          "tool_ref": "identity-tools",
          "human_approval": "NONE"
        }
      ],
      "evaluation_strategy": {
        "evaluation_strategy_id": "native",
        "max_steps": 15
      }
    },
    "config": {
      "provider_id": "openai",
      "params": {
        "model": "gpt-4o-mini",
        "temperature": 0.1,
        "api_key": {"$secretRef": "RESPONSE_PROVIDER_API_KEY"}
      }
    }
  }
}

Even though the spec sets human_approval: "NONE", the destructive operation reset_user_password always forces approval because of its severity. The agent with role: editor cannot run destructive operations at all; only an operator agent can.

Content shields on tool results

Diagnostic tools can return sensitive data. Attach shields to the tool registration to sanitize outputs before the model sees them.

{
  "registered_tool_id": "identity-tools",
  "provider_id": "mcp",
  "connection_config": {"url": {"$secretRef": "IDP_MCP_URL"}},
  "operations": [
    {"name": "check_user_status", "severity": "read-only", "tier_grants": ["reader", "editor", "operator"]},
    {"name": "unlock_user_account", "severity": "mutating-recoverable", "tier_grants": ["editor", "operator"], "approval_required": true},
    {"name": "reset_user_password", "severity": "destructive", "tier_grants": ["operator"]}
  ],
  "shields": {
    "pii-redaction": {
      "provider_id": "shield-config",
      "connector": {
        "provider_id": "alquimia",
        "profile": {
          "system_prompt": "If the tool result contains a password hash or raw token, respond with {\"contains_secrets\": true}. Otherwise {\"contains_secrets\": false}.",
          "evaluation_strategy": {
            "evaluation_strategy_id": "one-shoot",
            "structured_output": {
              "method": "json_schema",
              "json_schema": {
                "type": "object",
                "properties": {"contains_secrets": {"type": "boolean"}},
                "required": ["contains_secrets"]
              }
            }
          }
        },
        "config": {"model_ref": "gpt-4o-mini-classifier"}
      },
      "action": "block",
      "threshold": 0.8,
      "block_message": "Tool result redacted: sensitive credential data detected."
    }
  }
}

Knowledge and OCI promotion

Runbooks and known-error articles live in a registered topic. After updating a runbook, the platform team republishes the agentspace and pulls it into production, so every worker gets the latest tools, topics, and classifications atomically.

alquimia registry topics add runbooks \
  --severity read-only --tier-grants reader,editor,operator
alquimia registry topics add-file runbooks ./runbooks/app-lockout.md --hydrate

curl -X PUT "http://runtime:8080/registry/publish?agentspace_id=itops&tag=v1.3.0" \
  -H "Authorization: Bearer $API_TOKEN"

For application teams

Ticket-driven workflows

Agents can be triggered by new tickets, comments, or status changes and write back updates through the same integration.

Tool orchestration

Connect diagnostic scripts, cloud APIs, and configuration-management tools as reusable capabilities the agent can invoke.

Approval gates

Sensitive actions such as password resets or access grants pause for human approval through Slack, Email, or the ticket itself.

For platform / SRE teams

Least-privilege execution

Agent roles are scoped to the smallest set of tools and credentials needed for a ticket category.

Immutable audit trail

Every diagnostic result, approval, and action is recorded in the worklog for later review and post-incident analysis.

Controlled blast radius

Agents run in dedicated workers with restricted credentials, limiting the impact of a misbehaving tool or prompt.

Common integration points

System	Role
ITSM / ticketing	Receive requests and write back status
Identity provider	Verify users and manage group membership
Cloud APIs	Restart services, scale resources, check health
Monitoring	Correlate tickets with alerts and metrics

Example outcomes

Reduced mean time to resolve common incidents.
Fewer escalations to senior engineers for routine changes.
Consistent, auditable change control for compliance frameworks.