Skip to content
Work with agentsOperational playbooks

Make investigations repeatable

Capture investigation guidance, review a draft, and run a published workflow against an exact subject.

Availability: Operational playbooks must be enabled for your workspace. If the feature is unavailable, ask your workspace owner or ConvergeSense support about access.

Playbooks combine investigation guidance with approved capabilities. They help an agent gather consistent evidence while keeping missing data and uncertain conclusions visible.

Discover a playbook

Use convergesense_search_playbooks to find published guidance and templates. Fetch the selected playbook with convergesense_get_playbook to read its schema, capability coverage, and permitted actions.

Resolve the subject to an exact ticket or service identifier before execution. A playbook should not enumerate unrelated customers to make up for an ambiguous identity.

Create and review a draft

Owners can open Operational playbooks from Agent connections or Automations, adopt appropriate capability mappings, and create a draft from a template.

Edit the guidance and escalation instructions, inspect the changes, then save and validate the draft. Synthetic simulation makes no equipment calls and does not verify live connectivity.

An owner can also ask an external agent to prepare a draft:

text
Draft an investigation using our approved capabilities.
Validate it, save it, and give me the web review link.

Agent authoring requires current owner membership and explicit convergesense.admin consent. Existing read-only integrations are not upgraded automatically. Draft authoring tools cannot publish, schedule, or change tool policies.

Publish deliberately

Review and publish the revision in the web platform. Saving or publishing a playbook does not run or schedule it automatically.

An optional live draft test requires explicit authorization for one exact subject. It runs an immutable snapshot, so later edits do not change a run that already started.

Run and follow progress

Call convergesense_run_workflow with the workflow identifier supplied by the playbook. Poll convergesense_get_workflow_run using the original run ID, including after restarting your client.

Cancellation prevents future steps. It does not interrupt or replay a device command that was already transmitted.

Interpret the evidence

A completed collection is not a declaration that the circuit is healthy. Read the per-source facts, observation times, units, coverage gaps, identity evidence, and suggested next checks.

Keep advertised plan rates, device configuration, trained rates, and measured traffic distinct. Current health does not exclude an earlier intermittent fault, and counters are not automatically incident-window deltas.

Results expire 15 minutes after completion. Retrieving expired evidence does not silently rerun equipment.

Spotted something we can improve?Edit this page on GitHub