Skip to main content
Momentic uses several specialized AI agents. Each is versioned independently, so you can upgrade one without changing the others. Older versions stay available for backwards compatibility; set a specific version per agent in your momentic.config.yaml ai.agentConfig block. Momentic announces agent deprecations at least 90 days in advance.
When upgrading an agent version, we recommend testing the migration in a branch first. In rare cases, AI model decisions can vary from version to version.
The recommended agents use the latest models. They are tuned to use memory (insights saved from past runs) and custom knowledge base entries. All agents have built-in fallback mechanisms to handle AI provider outages and disruptions. During such downtime, agents may incur additional latency or exhibit different behavior. Memory, the knowledge base, and failure recovery reduce the frequency and impact of AI non-determinism on your tests.

​
Web agents

v5 latency is the median across all effort levels, weighted by production traffic.

​
Mobile agents

​
V5 effort routing

V5 requires momentic 3.52.10 or later. v5 routes each locator and assertion query to one of three effort levels (fast, medium, or extended) based on the query’s complexity. You set v5 in agentConfig; the effort level is not configurable. Fast effort has a median latency of 2.4s, 35% lower than the 3.7s median of v4. The effort level sets the credit usage for the query: Credit multipliers apply only to credit-based organizations.

​
What each agent does

​
Locates elements from a natural language description, used by Click, Type, and Element check steps. Prioritizes exact single-quoted text values, builds more accurate caches, and achieves 15% lower latency compared to v3. Scores 10% better on our internal web automation benchmark.
​
Evaluates natural language statements against a snapshot of the page, used by AI check steps. Prioritizes exact single-quoted text values and discerns visual detail more accurately. Scores 8% better on our internal web assertion benchmark with 22% reduced latency compared to v3.
​
Evaluates natural language statements from a viewport screenshot. Reasons about position more accurately than previous versions. Scores 12% better on our internal visual understanding benchmark.
​
Extracts structured data from the page given a JSON schema, used by AI extract steps. Supports nested objects and arrays.
​
Generates and executes recovery steps when a recoverable failure is detected. Requires ai.failureRecovery.
​
Locates elements in the Android UI hierarchy from a natural language description, used by mobile Tap, Type, and Element check steps.
​
Locates elements in the iOS UI hierarchy from a natural language description, used by mobile Tap, Type, and Element check steps.