Cup'n'String
Join Waitlist

© 2026 Cup'n'String

The Enterprise Security Brain for AI Agents

AI agents are powerful.
Give them boundaries.

A standalone security control plane that automatically discovers local runtimes, orchestrates host firewalls, secures model endpoints, and compiles compliance evidence for your developer workstations.

Join the waitlist Explore Capabilities

Model Context Protocol &
Tool Governance

Manage and inspect the tool configurations available to local agents. Prevent AI models from running rogue terminal processes, executing out-of-scope code deletes, or reading confidential workspace context.

Enforce strict tool schema and parameter boundary limits.
Dynamic prompt/response risk classification and PII redaction.
AI Tool Gating Flow
Real-Time
AI Client
(IDE/CLI)
Gateway 127.0.0.1
Local Gateway
Intercept
Policy Engine
Rule Matcher
(Inspect Tool)
INTERCEPT LOG active
✔ tool: fs_read_file("/src/index.ts") → APPROVED
✘ tool: shell_run("rm -rf /") → BLOCKED

Sovereign Firewall
Orchestration & Verification

Programs policies into native firewall rulesets rather than forcing complex custom overlays. Cup'n'String translates policy intents into platform-native pf, nftables, or SASE routes and continuously scans for config drift.

Generates clean change sets before applying security rules.
Tamper-evident audit trail backed by cryptographically signed HMAC evidence.
Firewall Orchestration
Orchestrated
macOS pf / Windows Defender Active

Device Operating Mode: MANAGED_AGENT

CHANGE SET PREVIEW 1 Proposed
@@ -23,8 +23,9 @@
- pass out proto tcp to any port 8080
+ block out proto tcp to any port 8080
+ pass out proto tcp to 127.0.0.1 port 8080
Evidence Hash Generated
HMAC Verified

Local Container &
Shadow IT Discovery

Inspect local workstation environments read-only. Detect running Docker containers, Compose setups, and Kubernetes services. Identify shadow AI engines (Ollama, LM Studio) and expose them securely using outbound reverse tunnels (gRPC/WebSockets).

Bypasses secrets, env vars, and code configs for strict read-only privacy.
Converts shadow containers to formal tenant-governed endpoints.
Local Service Scanner
Read-Only
Docker / Compose
3 containers found
Kubernetes Cluster
Kubeconfig Active
🔐 Privacy Guardrails Active: Zero collection of environment secrets, Kubernetes config secrets, database passwords, or raw inspector payloads. Scanner strictly lists container labels and exposed ports.

Zero-Trust API
Credential Shielding

Ensure developer workstations never store credentials for cloud LLM services. Requests are proxied locally, injecting API key references on the gateway, eliminating workstation leakage risks.

Stores keys in tenant-isolated encrypted keystores.
Tracks cost quotas and token limits per developer.
Credential Shielding
Zero-Trust
1. Workstation request No LLM Key
2. Provider Gateway Injects Key Reference
3. Outbound LLM Call Keys Injected
"Credential keys reside encrypted in the Tenant Secret Store, never on developer laptops."

Full Security Features Catalog

Dynamic catalog managed directly via our hidden file-based Markdown CMS.

Bo

MCP & Tool Governance

Manage and audit what tools and servers AI agents can access on developer workstations.

Collection: features
Sh

Firewall Orchestration

Program and verify host-level OS firewalls and corporate Zero Trust networks.

Collection: features
Se

Shadow AI Discovery

Discover unmanaged local AI engines and rogue developer tools.

Collection: features

Join the Waitlist

Be the first to secure your developer machines and govern AI agent runtimes.