Agents are a different workload than websites — they chain tool calls, wait on LLMs, and need to run close to whatever they're calling.
DIMENSION
HYPERSCALERS
CLOUDFLARE
Cold start
Lambda: 100–500ms · containers: seconds+
Workers: under 5ms
Billing model
Wall-clock time — billed while waiting
CPU time only — not billed while idle
Agent-service reach
Bedrock / Foundry: a handful of regions
330+ cities, edge-native
Security model
IAM + separate WAF/CDN vendor, bolted on
Zero Trust, Access, Gateway — same network
Honest caveat: hyperscalers still lead on deep managed ML training and GPU pipelines. Cloudflare's story is serving and securing agents at the edge — not training foundation models.
ARCHITECTURE, NOT ASSEMBLY
Purpose-built, not assembled from parts.
On traditional cloud, an agent means stitching together six services before writing any agent logic.
Compute service
Database
Message queue
Container runtime
Secrets manager
API gateway
On Cloudflare, it's one integrated framework:
Agents SDK
Durable Objects — state lives with the agent
Workflows — durable retries + human-in-the-loop
THE PRICING ARGUMENT
Pay for thinking, not for waiting.
Agents spend most of their time waiting — on the LLM, on a human approval, on a scheduled trigger.
Traditional VM / container alive 24 hours
24 hours
Cloudflare Durable Object alive 24 hours, hibernates when idle
10 seconds
An agent that runs 10 seconds of actual compute but stays "alive" for 24 hours costs 10 seconds — not a day of a server.
THE NEW PRIMITIVE · CODE MODE
Don't call tools — write code.
Models are trained to write code. So give the agent one code tool, not fifty — it writes a plan that runs in a sandbox and returns only what matters.
DIRECT TOOL CALLS
Every step round-trips the model
LLM→tool→LLM→tool→LLM→tool
CODE MODE
One program composes it all in-sandbox
LLM→code · many calls · one result
up to 80%fewer inference tokens & cost
// the model writes this — not 5 separate tool calls const projects = await codemode.list_projects({ status: "active" }); const blocked = []; for (const p of projects)
blocked.push(...await codemode.list_tasks({ projectId: p.id })); return blocked.filter(t => t.status === "blocked");
A Developer Platform primitive — code runs in an isolated Dynamic Worker, so credentials stay out of the model context. On MCP server portals, Code Mode is on by default.
05
Let's bring it to market together.
Cloudflare brings the platform. NTT DATA brings transformation and delivery expertise.
WHO DOES WHAT
Cloudflare provides the toolkit; you bring expertise & white-glove services.
Cloudflare's own
NTT DATA-led
Comprehensive security & cost controls
AI Gateway
Zero Trust Network Access
MCP server hosting + portals
Full-stack developer platform
Storage
Compute
AI
Starter kit of skill & context files
Cloudflare's own
Partner augmented
Consulting to build the context layer
Shape use cases, workflows, skills and enterprise context
Implementation & managed services
Customize experiences
Integrate enterprise systems
Operate and evolve
FOUNDATIONAL WORK
Invest in a central context layer.
Email Alias
Magic inbox for work requests
Expert Tasks
Documented expert workflows
↓
Triage & Audit
AI agent operators curate the work
↓
Skill Repo
Curated skill files + context, centrally managed
↓
AI Agent Workspace
Surfaced and ready-to-use in Cloudflare OS
A proposed NTT DATA consulting motion: define jobs-to-be-done as centrally managed skill files.
1
Document workflows
Tap experts to document workflows and create skill files.
2
Open a magic inbox
An email alias where everyone else can send work.
3
Staff it with agents
AI agent operators run the inbox as a service.
4
Curate into skills
Collect the use cases and convert them to skill files.
5
Surface in Cloudflare OS
Everything is ready-to-use for the whole team.
SKILLS VS CONTEXT
Two file types, one knowledge layer.
Skill File
Teaches an agent how to do one task — like a recipe.
When to use it — what triggers the task
Step-by-step, each step an action verb
Common mistakes to avoid + if/then logic
One page ideal · .md
Context File
Gives an agent background knowledge — like a glossary.
Overview — 2–3 sentences on the topic
Key definitions — terms, plain-language
Quick reference table — "if X, use Y"
One–two pages · .md
DEPLOYMENT MODEL · PROPOSED
From pilot to production in 10 weeks.
An illustrative joint delivery model — Cloudflare enables; NTT DATA leads transformation, integration and operations.
01
Design
4 WEEKS
Stakeholder mapping
Technical review
Project plan & customization
02
Build
4 WEEKS
Context & skills buildout
Cloudflare OS app deployment
MCP integration
Pilot deployment & launch
03
Manage
ONGOING
2 weeks post-launch support
Steady-state operations
Ongoing context updates
Proposed scope: NTT DATA leads architecture, security, identity & operations — not just code delivery. Confirm roles and timing during pilot design.
ENGAGEMENT ECONOMICS · ILLUSTRATIVE
The business case for both sides.
Reference model for a 5,000-seat digital-native client — partner services plus recurring Cloudflare consumption.
THE NUMBERS
~$1.7M
partner services revenue over the program
$650K
monthly recurring, at steady state
LLM token cost~$570K / mo
Cloudflare services~$80K / mo
DELIVERY TEAM
4
onsite FTE full program duration
6
remote FTE flexed across phases, not all full-time
Illustrative estimate. Figures are based on a 5,000-seat reference deployment from Cloudflare's internal experience — not an NTT DATA quote, forecast or approved offer. Actuals depend on scope, seat count & model choice, and are validated during the MVP.
06
Get Started
Standing up a trial environment, with provisioning agreed jointly.
SETUP · PART 1
From zero to running.
1
Confirm trial environment
Cloudflare and NTT DATA align entitlement, ownership and provisioning.