Home โ€บ Blog โ€บ The agent layer, mapped
Inside the Agent Layer ยท Part 1 ยท The map

The Agent Layer, Mapped: What Runs Between a Request and the 400th Tool Call

An AI agent looks simple from outside: you ask for something, and some minutes later it is done. Inside, a model has called tools hundreds of times, run code in isolated computers, opened browsers, read and written files, remembered what it learned, asked permission, recovered from crashes and spent many times the tokens of a chat. This series follows the infrastructure that makes that possible, the newest layer of the AI factory. Part 1 maps it end to end.

FL
FrontierAGI Team
โš ๏ธ How to read this (as of 28 Oct 2026) Token multipliers come from Anthropic's engineering write-up; MCP numbers from Anthropic's December 2025 ecosystem update; startup funding from press reports (marked โš ๏ธ where sources disagree). The request trace and token calculator are our illustrations. Disclosure: Claude, made by Anthropic, helps write this site; Anthropic created MCP and publishes some figures used here, which we cite like any other source.

1. Case file

๐Ÿค–
The agent layer, October 2026The top of the ML layer, seen from the infrastructure side
Tokens per task vs chat (Anthropic)Agents ~4ร—; multi-agent systems ~15ร—
MCP, Dec 2025 (Anthropic)10,000+ active public servers; 97M+ monthly SDK downloads
Agent sandbox start-upsDaytona $1B valuation; Browserbase $300M; E2B Series A (reported)
Gartner (Jun 2025)>40% of agentic AI projects cancelled by end of 2027 (forecast)
Performance driver (Anthropic research agent)Token usage explained ~80% of variance

2. The loop

Every agent, from a coding assistant to a customer-service bot, runs the same basic loop: the model reads the goal and the context, decides on an action (usually a tool call), the runtime executes it, the result goes back into the context, and the loop repeats until the model decides it is done or something stops it. Our earlier articles cover the loop as a design pattern (the agentic loop, harness engineering) and the frameworks that implement it (ML Layer Part 8). This series is about what it takes to run that loop millions of times a day, safely, for hours at a time.

3. The eight components

The agent layer, top to bottom

Tap a component for its job, typical tools and the part of this series that covers it.

4. One request, traced

"Fix the failing test in my repo and open a pull request"

A composite of how coding agents work. Press Play. Timings are illustrative.

Press Play.

5. Why agents are expensive

Anthropic reported that in its data, agents typically use about 4 times the tokens of a chat interaction, and multi-agent systems about 15 times. In its multi-agent research system, token usage alone explained about 80% of the variance in performance, which is why spending more tokens often buys better results, and why multi-agent designs only pay off for valuable tasks.

Price a task

Our model: chat-sized tokens ร— multiplier ร— price. Change the inputs.

Multiply that across an industry and agents become a major driver of inference demand, the shift our compute series calls the Great Flip. Part 7 covers how teams track and control agent costs.

6. Who builds what

Model labsShip models trained for tool use plus their own runtimes: OpenAI's Agents SDK and Responses API, Anthropic's Claude Agent SDK, Google's Agent Development Kit (see our harness comparison).
Sandbox and compute start-upsE2B, Daytona, Modal and others rent isolated machines for agents to run code; Browserbase and others rent browsers. Reported rounds: Browserbase $40M at a $300M valuation (Jun 2025), E2B a $21M Series A (Jul 2025), Daytona $60M at a $1B valuation โš ๏ธ.
Workflow and durabilityDurable-execution engines such as Temporal, and graph frameworks such as LangGraph, keep long agent runs alive across failures (Part 3).
Protocols and gatewaysMCP for tools and A2A for agent-to-agent communication, both now hosted by the Linux Foundation's Agentic AI Foundation; gateways from cloud and API-management vendors (Part 4).
Identity, memory and observabilityIdentity providers adding agent credentials (Part 5), memory stores (Part 6), and tracing and cost tools such as LangSmith, Langfuse and Arize (Part 7).

Gartner forecast in June 2025 that over 40% of agentic AI projects would be cancelled by the end of 2027 because of cost, unclear value and weak risk controls, and warned of "agent washing". Much of what separates a demo from a dependable agent is the infrastructure in this series.

7. Where it sits in the factory

On our site plan of an AI factory, agents run in Data Hall B alongside inference and evaluation: the model calls go to the inference fleet, while sandboxes, browsers and tool gateways run on ordinary CPU servers next to it. In the cutaway view, agents are the top floor. The same machinery also runs inside training: RL environments are agent sandboxes used to generate training data (Part 8).

8. The series

Inside the Agent Layer: eight parts

VerdictThe agent layer is where model capability turns into work done, and it is mostly ordinary distributed-systems engineering applied to an unusual workload: long, branching, expensive, partly unpredictable jobs that act on real systems. The model decides; the layer around it decides whether those decisions run safely, survive failures, remember the right things, stay within budget and can be audited afterwards. The next part starts at the bottom: the sandboxes where agents act.

9. Reading the numbers

Primary / widely confirmed
Agents ~4ร— chat tokens; multi-agent ~15ร—; token usage explains ~80% of variance; multi-agent beat single agent by 90%+ on Anthropic's internal research eval (Anthropic engineering, Jun 2025). MCP 10,000+ active public servers, 97M+ monthly SDK downloads (Anthropic, 9 Dec 2025). Gartner forecast (Jun 2025).
Reported โš ๏ธ
Start-up funding rounds and valuations (press and aggregator reports; E2B totals and Daytona dates differ across sources).
Our illustrations
Component list, request trace and timings, token calculator.