SolarWinds' Avior lets models choose IT fixes inside a fixed, auditable framework
SolarWinds launched Avior, a standalone AI agent for first-line IT work that holds anything it classes as potentially dangerous for human approval. The company says models decide what needs doing while a fixed, auditable framework controls how, so a rollout turns on where the vendor draws that line.
The Product Desk
SolarWinds classes which actions are potentially dangerous; those wait for human approval. Each action runs with its starter's permissions and is logged. Avior reaches third-party systems via the Model Context Protocol. Customer data is masked, not used for training.
- decision SolarWinds Classes which actions are potentially dangerous; those wait for explicit human approval, claim 8
- exposure Person who starts an action Each Avior action runs with this person's permissions and is logged, claim 9
- capability Third-party systems Reached through the Model Context Protocol, with no data migration needed, claim 6
- constraint Customer data Masked and never used to train the underlying models; only request-relevant data is sent, claim 16
| Who | How | Kind | Claim |
|---|---|---|---|
| SolarWinds | Classes which actions are potentially dangerous; those wait for explicit human approval | decision | 8 |
| Person who starts an action | Each Avior action runs with this person's permissions and is logged | exposure | 9 |
| Third-party systems | Reached through the Model Context Protocol, with no data migration needed | capability | 6 |
| Customer data | Masked and never used to train the underlying models; only request-relevant data is sent | constraint | 16 |
What happened
- Avior is a standalone application on the SolarWinds Platform and connects to third-party systems via the Model Context Protocol, with no data migration needed.
- Its launch jobs are merging related alerts into one incident, catching out-of-channel configuration changes and restoring last-known-good, and retiring records for devices long gone.
- Every action runs with the permissions of the person who started it and is logged.
- Only data relevant to a request is sent, and customer data is masked and never used to train the underlying models.
- Avior is available from launch day through a guided onboarding program.
Why it matters
- decision Whoever signs off on a rollout has to decide whether SolarWinds' definition of dangerous matches their own change policy, because that definition sets the size of the approval queue.
- exposure Actions inherit the starting person's permissions and can reach third-party systems, so over-scoped accounts become the widest thing the agent can touch.
- constraint Coverage today is three named jobs, and the models choose within goals SolarWinds has predefined, so the agent's reach is the list the vendor has written.
An alert fires for a device that left the network long ago, and someone has to work out that the record is stale. Retiring those records is one of the jobs Avior takes on [7]. We'd expect a tiered-autonomy agent to start with work like that, since a wrong call is easy to spot and rollback support is built in [10].
SolarWinds sells Avior on its architecture. Chief Technology Officer Krishna Sai said the design grounds "probabilistic models in deterministic execution" [3]. Interim CEO Cullen Childress, who is also chief product officer, said "reaction time alone is no longer enough" as environments grow more complex [11][12].
The thing being done is narrower than that sentence suggests. SolarWinds' position is that where uptime and security are at stake, an agent whose results vary between runs is a liability [2]. Avior's models still decide what needs doing, and a fixed framework of predefined goals and approval controls governs how it gets done [4]. In our view the run-to-run variation does not go away. It moves to the step where the model chooses what to do. Sai said those goals draw on 25 years of SolarWinds experience with how IT environments actually behave [5].
Autonomy in Avior is tiered by risk [8], so the line between acting alone and waiting for a person is SolarWinds' classification of what is potentially dangerous. SolarWinds did not disclose pricing [14], and the coverage does not say which actions land on the approval side or whether customers can move the line. The company said more capabilities are on the way [15].
What teams tell themselves is that an agent earns autonomy by being clever. What the tiers encode is the cost of a wrong call, and two questions sort any task on that basis: can the change be undone, and does it touch devices that matter most?
Reversible and low-stakes work can run unattended. We'd put merging related alerts into a single incident in that cell [7]. Reversible but high-stakes work, such as restoring last-known-good configuration on the devices that matter most [7], goes behind approval at first. Anything irreversible stays behind approval whatever the history says.
We'd open only the first cell on day one. The cost is an approval queue that stays long until the logs [9] show the agent's calls holding up on a given task, and a task moves to the faster cell only after that.
What to watch
- Whether SolarWinds publishes which actions sit in the approval tier and lets customers change the tiers.
- Pricing, and what the guided onboarding program asks of customers.
- Which capabilities SolarWinds adds beyond the three launch jobs, and how each one is tiered.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+30
- Incentives60
- Confidence65
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
SolarWinds unveiled Avior, a standalone agentic AI system built to take on first-line operations work for enterprise IT teams.
- [2]
SolarWinds argues that an agent producing different outcomes from one run to the next is a liability wherever uptime and security are on the line.
- [3]
Chief Technology Officer Krishna Sai said the design grounds "probabilistic models in deterministic execution."
- [4]
The models work out what needs doing; a fixed framework of predefined goals and approval controls governs how it gets done, and every workflow can be audited.
- [5]
Sai said the predefined goals draw on 25 years of SolarWinds experience with how IT environments actually behave.
- [6]
Avior runs as its own application on the SolarWinds Platform and reaches third-party systems through the Model Context Protocol, with no data migration needed to use it across self-hosted and cloud environments.
- [7]
At launch Avior merges related alerts into a single incident, spots configuration changes made outside normal channels and restores the last-known-good configuration on the devices that matter most, and retires records for devices long gone from the network that still fire alerts.
- [8]
Autonomy in Avior is tiered by risk, so anything SolarWinds classes as potentially dangerous waits for explicit human approval.
- [9]
Each Avior action runs with the permissions of the person who started it and is logged.
- [10]
Rollback support is built into Avior.
- [11]
Cullen Childress, who is also SolarWinds' chief product officer, became interim chief executive when Sudhakar Ramakrishna's tenure ended Sept. 30.
- [12]
Childress said "reaction time alone is no longer enough" as environments grow more complex.
- [13]
Avior is available starting launch day through a guided onboarding program.
- [14]
SolarWinds did not disclose pricing for Avior.
- [15]
SolarWinds said more Avior capabilities are on the way.
- [16]
Only data relevant to a request is sent by Avior, and customer data is masked and never used to train the underlying models.
- [17]
Avior has three named launch jobs: alert merging, out-of-channel configuration change detection with restore, and stale inventory retirement.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
2 independent publishers whose own reporting we read for this story.
- helpnetsecurity.comSolarWinds Avior pairs IT automation with human approval and audit trails
1 article · October 14, 2026
- siliconangle.comSolarWinds’ Avior pairs AI agents with deterministic execution for enterprise IT
1 article · October 14, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- IT operations automationFollow
- Agentic AI and tool useFollow
- Model Context ProtocolFollow
- ObservabilityFollow
Entities
- Model Context ProtocolFollow
- SolarWindsFollow
- AviorFollow
- Radiology PartnersFollow
- Cullen ChildressFollow
- Krishna SaiFollow