Automation Platform

Sign in to view the User Guide

This guide is available from the authenticated platform workspace.

Return to sign in
Automation PlatformUser Guide

Operator edition · English

User Guide

A practical guide to observing the network, planning safe changes, approving work, executing through governed workers, and reviewing evidence.

Reviewed 01 Sep 2026 Role-aware operations Printable reference
First session
  1. Check the top bar.Confirm Device I/O, Workers, and PG status.
  2. Open Topology.Review reachability and the latest link evidence.
  3. Open Network Elements.Confirm access readiness before any operation.
  4. Use your assigned role.Plan, approve, execute, or observe only what you are authorized to do.

Core operating model

Use the same safe workflow every time

Planning and approval do not change a device. Device access begins only when the required gates pass and a worker claims authorized work.

Governed change lifecycleShort animation · no device action
01ObserveFresh evidence
02PlanTyped intent
03ValidatePre-check + diff
04ApproveFrozen scope
05ExecuteWorker + lock
06VerifyEvidence + state
Before

Confirm scope

Use exact devices, fresh inventory, collected resources, and the intended execution path. Read the full preview.

During

Follow state

Do not repeat a request while it is queued or running. Open its detail view and follow the evidence.

After

Verify outcome

Check configuration presence and operational state separately. Create a governed withdrawal when the resource is no longer needed.

Stop on uncertainty.

If a run shows unknown_outcome, do not retry. The platform keeps the device write lock. A Platform Administrator must verify the device out of band and record the result.

Global controls

Read the app shell before opening a workflow

The top bar and left status card show the platform state that applies to every page.

01

Global Search

Press /, then search by network element, management IP, device, serial number, software version, or job. Use arrow keys and Enter to open a result; Esc closes the list.

02

Device I/O

READ-ONLY means device I/O is globally off. DEVICE I/O ON means access may proceed only if every other gate also passes.

03

WORKERS

A green dot means all required runtime services are available. A red mark needs attention. Click the badge for service, heartbeat, queue, and Start or Restart controls.

04

PG status

PG OK confirms the database health check. Treat PG DOWN as a platform incident. Do not start new work.

05

Theme and account

Theme preference is saved in this browser. Open the account menu or Users & Access to review your profile, group, effective permissions, password controls, and sign out.

06

Permission-aware UI

Navigation, management panels, and operation buttons appear only when your effective permissions allow them. Users & Access stays available for your own account. A hidden control is not a platform failure; check your assigned group or overrides.

Starting a worker may start real work.

Before Start or Restart, inspect its queue depth and purpose. When Device I/O is on, an available worker may immediately claim already-authorized items.

Overview

Topology

Use the topology for network context, link evidence, traffic context, and fast access to a device inspector.

Common operations

  1. 1
    Review the legend.

    The node ring shows role; the center shows reachability. Link color shows operational state and line style shows confidence.

  2. 2
    Move around the canvas.

    Drag nodes to reposition them and scroll to zoom. Use Reset Layout to return to the computed placement.

  3. 3
    Choose labels.

    Show device name, system IPv4, or system IPv6. Keep labels limited when the view is dense.

  4. 4
    Open a device.

    Select a node to open Inventory, Configuration, Readiness, Attributes, Allowlist, and Auto-Init details.

  5. 5
    Use contextual actions.

    Run allowed read operations, open traffic evidence, or use the permission-gated link handle to begin a LAG plan between two devices.

Overview

Dashboard

Use the Dashboard as the platform summary, not as proof that one device is ready.

What to review

  • Network element and device totals
  • Estate-wide active and failed Discovery Jobs
  • Recent job duration and status
  • Safety status, database, and access gates
  • Credential Profiles only when you have credentials.manage; otherwise your access level

Recommended response

  • Open any failed or long-running job
  • Check stale evidence before acting
  • Resolve platform health alerts first
  • Use Network Elements for device-level truth

Network

Network Elements

This page owns the platform identity, reachability, readiness, trust, and management lifecycle of each network element.

ImportChoose OSVerify identityInitialize or adoptSyncOperate

Onboard and synchronize

  1. 1
    Import NEs.

    Add the approved inventory list. Check hostname, management IP, expected network OS, role, and tags before import.

  2. 2
    Choose the correct management path.

    Use Nokia SR OS Auto-Init only for the supported Nokia lifecycle. Use Adopt existing management for an already-configured Cisco IOS XR device.

  3. 3
    Verify device identity.

    Independently verify the presented SSH algorithm and SHA-256 host key. Never reuse trust from another address or accept a changed identity from the UI alone.

  4. 4
    Initialize or adopt.

    For Nokia, provide the approved bootstrap credential and reason. For Cisco IOS XR, choose a compatible stored read profile or enter a one-time login, add the reason, then select Adopt & Discover.

  5. 5
    Follow evidence.

    Open the Discovery Job. Cisco adoption must end as MANAGED · READ ONLY; Nokia must complete its supported management verification before normal Sync and automation.

  6. 6
    Open the inspector.

    Review Inventory, Configuration, Readiness evidence, Attributes, Allowlist, and lifecycle controls before planning a change.

Identity mismatch is a hard stop.

Do not confirm a changed hostname or host key from the UI alone. Verify through an independent source such as console, VM identity, serial, or approved device records.

Vendor boundaries are explicit.

Cisco IOS XR adoption sends only the fixed read-only discovery command set. It does not run Nokia Day-0, enter configuration mode, provide Cisco writes, or claim full readiness.

Controlled Re-enroll.

Re-enroll rechecks the host key, rotates managed CLI, NETCONF, and SNMPv3 credentials, rebinds profiles, and restores automation only after live verification.

Network

Devices & Inventory

Review facts collected by discovery: hardware identity, software, capabilities, configuration mode, ports, cards, routing, and services.

How to use it

  1. Select a device row.
  2. Check model, network OS, serial, and configuration mode.
  3. Open capability and inventory groups.
  4. Use service or routing detail records to start a contextual plan only when that network OS supports it.
  5. Open matching automation history to review provenance.

Read evidence correctly

  • Collected data is a timestamped snapshot.
  • Stale data is not a green result.
  • Configured does not mean operationally up.
  • An absent stanza may be a default omitted by running configuration.
  • Refresh evidence before using a resource selector.

Operations

Automation

Start with the desired network outcome. The platform builds typed, immutable per-target plans and shows exact configuration, checks, rollback, and hashes before execution.

ServicesVPLS, Epipe/VPWS, and VPRN over EVPN and SRv6
SR PolicyHead-ends, color, endpoint, preference, and segment list
LAGOne local or two-ended aggregate with collected members
Router InterfaceTwo-ended IPv4/IPv6 links over ports or LAGs
IGPIS-IS or OSPF attachment with symmetric intent
BGPCreate, reuse, add, edit, or remove groups and neighbors

Standard planning flow

  1. 1
    Choose type

    Select the exact domain or service.

  2. 2
    Select targets

    Use collected, available resources.

  3. 3
    Review plan

    Check every target payload, rollback, verification, and hash.

  4. 4
    Choose route

    Direct, approval, or MOP only when policy offers it.

  5. 5
    Follow evidence

    Track all targets until verified or stopped.

Multi-endpoint services

Select every participating NE, a collected port or LAG and VLAN for each endpoint, then set the common EVPN values. VPRN routing is configured per endpoint. Review each Nokia YANG payload in the matching endpoint card.

Lifecycle and withdrawal

A successful creation record remains immutable. When the resource is no longer needed, open it and create a governed withdrawal. Remove dependent routing before its parent service, or use the reverse MOP offered by a completed blueprint.

Direct does not mean unrestricted.

Direct mode may skip Change Request approval or MOP orchestration, but it still requires authentication, permission, Device I/O, allowlist, host-key pin, write credentials, validation, write lock, and audit evidence.

Configured and operational are different results.

A verified service or routing object proves the expected configuration is present. Check service, interface, adjacency, session, and traffic state separately when operational readiness matters.

Operations

Discovery Jobs

Review durable discovery execution history, active work, target scope, and append-only evidence.

Follow execution

  1. Open a job row to view metadata and its append-only audit timeline.
  2. Follow queued, running, partial, completed, failed, or cancelled state.
  3. Open failed evidence and record the error code before retrying.
  4. Cancel only a non-terminal job and confirm the target.

Read the totals

  1. Use Active for current worker load.
  2. Use Completed / Partial to separate full and incomplete collection.
  3. Use Failed / 24h for recent operational attention.
  4. Use the table for recent rows; estate-wide totals are not limited to that page.
A queued job may be waiting for authorization, not capacity.

If the detail says authorization is missing or expired, start a new authorized operation from the correct page. Do not start extra workers as a workaround.

Assurance

Readiness Matrix

Review the latest per-device decision from the onboarding readiness rule engine. This is evidence coverage, not generic continuous monitoring.

Review decisions

  1. Find the NE row and check each readiness column.
  2. Select a cell for Expected, Observed, Decision, and Evidence.
  3. Check the evidence timestamp and coverage.
  4. Treat missing, stale, warning, and fail as unresolved.

Use before change

  1. Confirm the NE is active.
  2. Confirm the required management planes are current.
  3. Open the source job or snapshot when evidence is incomplete.
  4. Refresh evidence before approving automation.

Assurance

Snapshots

Preserve normalized, checksummed configuration evidence and trace how each snapshot is used by governed workflows.

  1. Select a network element.
  2. Choose Capture Config if you have inventory permission.
  3. Wait for the new append-only snapshot.
  4. Select exactly two rows and choose Compare.
  5. Review additions, removals, source, size, collector, and checksum.
  6. Use Trace usage to find Change Requests, automation baselines, backups, or restores that reference the snapshot.
Failed collection is non-destructive.

A failed capture does not overwrite the last known-good snapshot. Always compare timestamps before using an older row as current evidence.

Operations

Manual Sync Runs

Run an operator-triggered, bounded-concurrency collection of readiness, inventory, and configuration across authorized NEs.

  1. Choose concurrency from 1 to 8.
  2. Select Run Manual Sync.
  3. Open the run detail.
  4. Review readiness, inventory, config, and errors for every NE.
  5. Resolve incomplete targets individually.
This is not a background production schedule.

The page records manual fleet runs. It does not claim that an unattended scheduler is configured.

Assurance

Fleet Readiness Report

Export portable fleet readiness evidence for review, handoff, or an approved maintenance record.

  1. 1
    Review the fleet table.

    Check reachability, onboarding, readiness time, snapshot, and access gates.

  2. 2
    Choose a format.

    HTML is the readable report; CSV supports analysis; JSON preserves structured data.

  3. 3
    Verify the export.

    Open the downloaded file and confirm its generation time and expected device scope.

Operations

Configuration Backup & Restore

Create verified local packages and use the governed candidate-based restore path. A restore does not directly overwrite config.cfg or require a reboot.

Backup
  1. Select devices.Use individual, multi-select, or select all.
  2. Choose the source.Running configuration via MD-CLI is the default. Startup config.cfg via SCP may differ from running state.
  3. Queue the batch.Each device gets its own durable task; one failure does not hide other successes.
  4. Verify integrity.Confirm state, source, byte size, timestamp, and SHA-256 manifest.
Restore
  1. Select a verified backup.Full replace is recommended; merge keeps extra live objects.
  2. Run preflight.The worker stages, loads, validates, compares, discards, and stores the diff.
  3. Review and approve.A Platform Administrator reviews the exact diff and enters the required batch phrase.
  4. Execute and verify.The worker rechecks baseline and diff, uses commit-confirmed, verifies post-state, then saves.

Automatic backups

Select devices, choose New schedule from selection, then set Daily, Every N days, or Weekly, the local time, IANA timezone, source, and enabled or paused state. Editing or deleting a schedule never deletes its existing packages or audit history.

Restore approval is high risk.

Check the device, backup source, artifact checksum, preflight baseline, complete diff, and mode. If the baseline changes after preflight, execution must stop. Never approve an unexplained diff.

Startup and running sources are not equivalent.

A startup SCP package reflects the saved BOF file. Unsaved running changes are not included.

Changes

Change Requests

Use a Change Request for controlled configuration with pre-check evidence, an exact dry-run diff, approval, separate execution, post-check, and rollback evidence.

DraftPre-checkDry-runApprovalExecute / Verify
  1. 1
    Create the draft.

    Select an approved template, target NE, and valid parameters. Review the frozen intent.

  2. 2
    Run Pre-check.

    The platform captures fresh configuration, checks access and drift, and records current values for rollback.

  3. 3
    Run Dry-run.

    Review the private-candidate diff. The candidate is discarded. An empty diff completes as a verified no-op.

  4. 4
    Submit and approve.

    The approver reviews the exact hash and diff. Approval issues a one-use authorization with a limited lifetime.

  5. 5
    Execute with another identity.

    The executor must differ from the approver. The worker reacquires evidence, matches the approved diff, commits, and runs post-check.

  6. 6
    Close the lifecycle.

    Confirm verified or rolled back. Use a typed withdrawal for a standing resource. Resolve uncertain outcomes only with external evidence.

Approval is bound to exact content.

If parameters, device baseline, or the live candidate diff changes, the approved authorization cannot be reused. Build and approve a new plan.

Changes

Template Library

Templates are vendor-scoped, versioned definitions that turn typed parameters into reviewed configuration and rollback plans.

Review, usage, and preview

  1. Search the Template Library and select a row.
  2. Review category, network OS, parameters, and version.
  3. Check exact saved-template links separately from renderer-level Change Request, Automation Target, and MOP Step references.
  4. Review the last-used time before revising or deleting.
  5. Render a placeholder preview and copy it only for approved review use.

Create or revise

  1. Choose Template types or New template.
  2. Use strict parameter types and patterns.
  3. Include forward checks and exact rollback.
  4. Review validation results and execution readiness.
  5. Save a new version; do not silently change approved intent.
Usage evidence has two sources.

A direct saved-template link and a renderer identifier are different provenance. Review both totals. The server blocks deletion when a governed reference still depends on the record.

Preview is offline.

A rendered preview does not connect to a device or push configuration. Secrets remain placeholders and must never be embedded in a template.

Operations

MOP Workflows

Use a Method of Procedure for a dependency-aware 1–N runbook with frozen plans, ordered execution, verification, stop conditions, and an independent reverse workflow.

1Build sequence

Choose a preset or arrange LAG, interface, IGP, BGP, VPRN, and other native components.

2Configure 1–N

Complete every component. Later steps may use projected results from earlier steps for planning only.

3Review all plans

Compile every child, then review target payloads, checks, rollback, dependencies, and immutable hashes.

4Govern

Set title and Minor, Major, or Critical tier. Critical work requires a maintenance window.

AI Assist and Provider configuration

Draft with an enabled Provider

  1. Open AI Assist when you have MOP planning permission.
  2. Select an enabled Provider and describe the operational goal without secrets.
  3. Review the sanitised draft and every system validation result.
  4. Move only accepted content into the native composer.
  5. Compile and review the same frozen plans as a manually authored MOP.

Configure Providers

  1. Confirm you have ai.configure.
  2. Use Provider configurations for OpenAI, Anthropic, or a compatible third-party service.
  3. Set name, protocol, Base URL, model, enabled state, and API key.
  4. Save, then use Test before operational drafting.
  5. Remove an unused profile only after checking impact.
AI configuration is a separate permission.

mop.plan permits draft use with the minimal enabled-Provider list. Only ai.configure exposes Provider URLs, model and key controls. API keys go directly to the encrypted broker and are never shown again.

AI output is advice.

AI cannot approve, execute, change hashes, weaken gates, or replace operator review. Never send device passwords, private configuration, customer data, or other secrets in a prompt.

After creation

  1. Review the MOP detail and all target steps.
  2. Collect the required signatures. The executor must differ from every applicable approver.
  3. Select Proceed / Execute once. The MOP Worker dispatches native children in strict order.
  4. Follow each child until verified. Any non-verified child stops later work.
  5. Export Excel for operator review, Markdown when required, or HTML as execution evidence.
  6. Use AI Review / Analysis only as advice. It cannot change frozen plans, hashes, approvals, or execution.
  7. For a completed blueprint, create a new governed withdrawal or recovery MOP when removal is required.

Remove completed history from the operational list

  1. A Platform Administrator selects one or more terminal runs, or chooses Clear completed history.
  2. Review the warning and scope. Active workflows remain disabled and cannot be selected.
  3. Enter an administrative reason of at least 8 characters.
  4. Type the displayed phrase exactly: ARCHIVE N MOP RUNS for a selection or CLEAR TERMINAL MOP HISTORY for all terminal history.
  5. Confirm once, then review the resulting audit event.
Delete history is a soft archive.

It hides terminal runs from the operational list but retains steps, approvals, child evidence, report hashes, and the append-only Audit Log. It does not delete or withdraw device configuration.

Do not assume automatic rollback.

A failed child stops later components. The platform does not blindly undo previously verified components. Use the explicit, newly planned reverse workflow when it is safe and approved.

Governance

Audit Log

Use the append-only audit log to reconstruct who did what, when, to which resource, and with what outcome.

Review sequence

  1. Filter by time, event, actor, or resource.
  2. Open the related job, change, service, automation, MOP, user, or runtime record.
  3. Match intent, payload, diff, artifact, and checksum identifiers.
  4. Record the error code and final resolution.

Audit hygiene

  • Never expect passwords or configuration bodies in audit rows.
  • Do not edit historical creation records after withdrawal.
  • Use timestamps and actor identity from the platform.
  • Keep external incident notes linked by approved reference, not pasted secrets.

Governance

Settings & Runtime Workers

Platform Administrators control device I/O, collection cadence, credential references, trust pins, relocation, and host-native services here.

Real-device I/O

Keep off during setup, migration, testing, or an unresolved incident. Enabling it does not bypass any per-device gate.

Traffic collection

Set the polling behavior appropriate for platform capacity and evidence needs.

Credential Profiles

Enroll a profile name, username, password, and purpose. Secrets go to the host-local broker and are never shown again.

Allowlist & fingerprints

Record only independently verified SSH SHA-256 fingerprints. Revoke trust when device identity is uncertain.

Runtime Services

Open WORKERS, inspect heartbeat and queue state, then start in dependency order beginning with Credential Broker.

Platform Relocation

Use only during a planned rebuild. Full estate retirement requires Device I/O off, settled work, and the exact confirmation phrase.

Credential and relocation actions are not routine fixes.

Do not replace a shared profile, reset the broker, retire inventory, or renew a host key to clear an unrelated error. Identify the exact failed gate first.

Governance

Users & Access

Every signed-in user can manage their own account. Four duty groups define the normal workflow. A Platform Administrator may edit non-admin group defaults and add or deny fixed delegatable permissions for one user.

GroupPrimary useTypical actions
Read-onlyObserve platform and network evidenceDashboard, topology, inventory, jobs, reports
PlannerCreate governed workInventory operations, changes, Services, MOP plans
Reviewer & executorReview and release workApprove, reject, execute, rollback, audit read
Platform administratorPlatform ownership and break-glass controlsUsers, governance, lifecycle, restore, workers, unrestricted terminal

My account workflow

  1. Open Users & Access from the sidebar or choose Profile & access from the account menu.
  2. Review your duty group, effective permissions, and any explicit removals.
  3. For a local platform account, update your display name or optional email and choose Save profile.
  4. Choose Change password to update your password. Use at least 7 characters with uppercase, lowercase, and a non-space special character.
  5. A password change signs out every session; a profile-only update keeps this session active.
  6. Ask a Platform Administrator to change your username, group, account status, or permission overrides.

Group defaults workflow

  1. Open Group defaults and choose Edit defaults for Read-only, Planner, or Reviewer & executor. Platform administrator is immutable.
  2. Keep platform.read; it is mandatory.
  3. Add or remove only the listed delegatable operational permissions.
  4. Treat ai.configure as sensitive and default-off. It may be granted to a group or one user when required.
  5. Do not look for terminal.admin here. Administrator SSH is available only as a special per-user permission.
  6. Review the final set before saving. Saving writes an audit event and revokes active sessions for current group members.

Administrator workflow

  1. Create the user with username, display name, optional email, duty group, and temporary password.
  2. Set individual permissions to Inherit, Add, or Remove only when a clear business need exists.
  3. Use terminal.admin only for a named user who needs audited Administrator SSH. It can never be a group default.
  4. Review the Effective permissions summary before saving.
  5. Tell the user that group, status, permission, password reset, or password change revokes existing sessions.
  6. Review the audit event after every access change.
A group does not remove separation of duties.

Even when one group has both review and execution permissions, the same signed-in person cannot approve and execute the same governed scope.

Administrator SSH is not a device credential.

The permission only reveals the governed terminal control. Exact target trust, allowlist, host key, device AAA, credential profile, Device I/O, and terminal authorization still apply.

Governance

Roadmap

Use this page to understand delivery status and planned capabilities. It is a product status view, not an operational control.

Complete

Implemented baseline

The application capability has current repository tests or witnessed evidence. Deployment-specific SSO, TLS, HA, DR, secrets, or scale gates may still remain.

Operational foundation

Useful but bounded

A real slice exists, but vendor breadth, live canaries, scheduling, or environment closure is still open. Read the stated boundary before use.

Planned

Direction only

The capability is not an available control. Do not infer an API, worker, schedule, or device action from a roadmap row.

Check current behavior in the product.

A roadmap item may be planned, partial, or complete. For an operational decision, rely on current UI controls, current evidence, and the active design contract.

Reference

Troubleshooting

Start with the failed gate. Avoid broad restarts or repeated submissions.

Buttons or pages are missing

Open the account menu and check group and effective permissions. Sign in again if access was recently changed because permission updates revoke existing sessions.

AI Provider controls are missing

MOP planning permission allows use of enabled Providers but does not reveal their configuration. Check whether your effective permissions include ai.configure. It is default-off outside Platform Administrator.

WORKERS shows attention

Click WORKERS. Check the exact service, heartbeat, installation state, and queue depth. Start or restart only the failed fixed service and only after checking pending authorized work.

Device is reachable but an operation is blocked

Reachability is only one signal. Check active lifecycle state, Device I/O, allowlist expiry, SSH host-key match, credential profile, readiness, and operation permission.

A newly imported Cisco IOS XR device is NOT ONBOARDED

Open the Network Element, select Cisco IOS XR · adopt existing management, independently verify the exact host key, choose a compatible read credential or enter a one-time login, add the reason, and use Adopt & Discover. Do not run Nokia Auto-Init.

Sync is INCOMPLETE

Open the per-NE detail. Identify whether Readiness, Configuration, or Discovery lacks evidence. Fix that exact dependency, then run the per-device Sync.

A plan is rejected for stale inventory or baseline drift

Collect fresh evidence, review what changed, and build a new plan. Never reuse an old approval or weaken the stale-evidence gate.

A worker run is queued for too long

Check authorization expiry, worker health, queue order, target write locks, and active unknown outcomes. Do not submit duplicates.

A run reports unknown_outcome

Stop all writes to the target. Do not retry, roll back blindly, or release the lock. A Platform Administrator must verify the live device through an approved out-of-band path and record verified or rolled back with evidence.

A completed MOP no longer appears

A Platform Administrator may have soft-archived terminal history. Check the append-only Audit Log and retained execution evidence. Archiving the list row does not withdraw device configuration.

The tutorial opens a sign-in message

Your platform session is missing, expired, or was revoked. Return to the platform, sign in, complete any required password change, then open Tutorial again.

Reference

Status glossary

Use state text and evidence, not color alone.

StatusMeaningOperator response
READY / VERIFIEDRequired checks or expected configuration evidence passed.Check freshness and operational state before the next dependent action.
QUEUED / RUNNINGWork is waiting or being processed.Open the detail view. Do not submit a duplicate.
INCOMPLETE / PARTIALSome work completed, but required evidence or targets are unresolved.Review per-step or per-target results.
STALEThe evidence is older than the accepted collection window.Refresh or recollect before using it for a plan.
FAILED / DOWNA deterministic error or unavailable dependency stopped work.Record the error code, fix the exact cause, then create a fresh attempt.
UNKNOWN OUTCOMEA connection was lost where commit state may be uncertain.Stop, preserve the lock, and perform approved manual verification.
UNENROLLEDThe record is historical and excluded from platform automation.Use controlled Re-enroll; do not renew trust directly.
NOT ONBOARDEDThe imported NE has no completed platform trust and read-management path.Choose the correct network OS and use its supported initialization or adoption flow.
MANAGED · READ ONLYCisco IOS XR fixed-command discovery passed with exact identity and authorization.Use collected evidence only. Do not assume write, Day-0, or full readiness support.
READ-ONLYGlobal real-device I/O is disabled.Planning and review may continue. Device collection or writes cannot.
Automation Platform User GuideUse current platform evidence and your approved operating procedure for production decisions.
Back to top