Ch 10 — MCP in Production & the Ecosystem

Deploying, scaling, and the growing world of MCP servers
IndexUnder the Hood →
High Level
-
Click play or press Space to begin...
Step- / 8
ADeployment ModelsLocal, remote, and hybrid approaches
1
computer
Local (stdio)Host spawns process
zero config
or
cloud
Remote (HTTP)Deployed service
shared by teams
or
lan
HybridLocal proxy to
remote backends
2
arrow_downward Packaging and distribution
BPackaging & DistributionHow servers reach users
package_2
PyPI / npmpip install or
npx to run
or
deployed_code
DockerContainer image
for isolation
or
link
Remote URLJust a URL
nothing to install
3
arrow_downward The server ecosystem
CThe Server EcosystemHundreds of servers across every domain
storage
DatabasesPostgres, SQLite
Redis, MongoDB
+
api
APIsGitHub, Slack
Jira, Stripe
+
4
folder_open
File SystemsLocal files, S3
Google Drive
+
build
Dev ToolsGit, Docker
Kubernetes, CI/CD
5
arrow_downward Hosts that support MCP
DHost EcosystemWhere MCP servers run today
smart_toy
Claude DesktopAnthropic's native
MCP host
+
code
CursorAI code editor
with MCP support
+
extension
VS Code + OthersWindsurf, Zed
Cline, Continue
6
arrow_downward Production best practices
EProduction Best PracticesMaking servers reliable and secure
security
Least PrivilegeMinimal permissions
scoped tokens
+
monitoring
ObservabilityLogging, metrics
error tracking
+
7
speed
PerformanceTimeouts, caching
rate limiting
+
verified_user
Input ValidationSanitize all inputs
prevent injection
8
arrow_downward The future of MCP
FThe Road AheadWhere MCP is going next
hub
RegistryCentralized server
discovery
+
swap_horiz
Agent-to-AgentServers calling
other servers
+
public
Open StandardIndustry-wide
adoption
1
Detail