Agent stack
The software layer that lets AI use tools and take multi-step actions.
Ranks moderate (60/100) — set apart by rising momentum (82) and cross-layer reach (74).
Why it matters
It is the bridge from models to autonomous application workflows.
Why now
Agentic workloads are scaling per-task compute sharply, intensifying inference demand.
If Agent stack runs short
Unbounded agent calls explode cost and latency.
In depth · editorial + model
The agent stack is the software layer that lets a model do more than answer — it lets the model call tools, query databases, retrieve documents and chain several steps together to complete a task. It is structurally important because it is the bridge from a raw model to a working application: without it a model is a text box, and with it the same model becomes something that can act. It matters now because agentic workloads run many model calls per task rather than one, sharply raising the compute each request consumes.
If costs run away, the danger is unbounded loops — an agent that keeps calling itself, multiplying inference spend and latency until the workflow stops paying for itself. The exposed names are the memory and retrieval layers agents lean on to stay grounded: vector database Pinecone, MongoDB and Elastic, all of which sit between the model and the data it reaches for. Their fortunes rise with agent adoption, because every extra step an agent takes is another read against the stores they run.
The companies exposed to Agent stack
How to think about it
- Agents multiply inference demand
What to watch
- Calls per agent task
- Agent reliability and cost control
Frequently asked
What is Agent stack?
The software layer that lets AI use tools and take multi-step actions.
Why does Agent stack matter for AI?
It is the bridge from models to autonomous application workflows.
Who makes Agent stack?
The companies the model tags as producers or suppliers of Agent stack: Datadog, Inc., Cloudflare, Inc., Pinecone, MongoDB, Inc., Elastic N.V..
Which companies are most exposed to Agent stack?
Datadog, Inc., Cloudflare, Inc., Pinecone, MongoDB, Inc., Elastic N.V. — 5 companies in total are mapped to Agent stack.
What happens if Agent stack runs short?
Unbounded agent calls explode cost and latency.
Where does Agent stack sit in the AI value chain?
Agent stack sits in the Models layer of the AI value chain.
Go deeper on Agent stack
- The materials, geographies and policies it depends on — heat-mapped
- Substitutes, relief valves and the domino chains if it tightens
- The live tension score, momentum and news drivers
- Four levels of analysis — from plain-English to strategic
model v0.7.0 · research, not advice
Model scores are illustrative reads from our model of the AI value chain — not investment advice.